DE102020209200A1 - Method for providing satisfaction information about a customer's predicted satisfaction with regard to a medical device - Google Patents

Abstract

The invention relates to a computer-implemented method for providing satisfaction information about a customer's predicted satisfaction with regard to a medical device. The method includes the method step of providing (PROV-01) input data, the input data including at least one operating parameter of the medical device and at least one piece of customer information. The method also includes the step of applying (APP) a first trained function to the input data, thereby generating the satisfaction information. The method also includes the step of providing (PROV-02) the satisfaction information.

Description

In the area of customer service, there are often many customer inquiries to be processed at the same time. In particular, a request can be a call and/or a ticket from a customer, for example. In particular, the customer can ask a question about how a medical device works, for example, or he can report a failure of a medical device or he can report a fault in the medical device, etc. In this context, it is often necessary to prioritize customer inquiries . For example, a customer who only has a question about a specific operation of the medical device can wait longer for a response from customer service than a customer whose medical device has completely failed. In particular, customers with serious problems should be given preferential treatment. Customers with frequently occurring problems should also be given preferential treatment. In particular, customer satisfaction should be ensured by prioritizing inquiries. In other words, it should be ensured that all customers receive the best possible care and support. This is to ensure customer satisfaction.

In addition, in customer service, it is often necessary to take appropriate action in response to a customer request. For example, customer service may decide that a call should be made to the customer. Alternatively, customer service can send a service technician to the customer. The action must be decided based on the request.

Prioritizing the customer inquiries or determining the appropriate action in response to a customer inquiry can be based in particular on information about the customer's satisfaction. It is known to determine customer satisfaction based on customer surveys or posts in social media using natural language processing (Graebner et al., "Classification of Customer Reviews based on Sentiment Analysis", 19th Conference on Information and Communication Technologies in Tourism , 2012; Bagheri et al., "Care more about customers: Unsupervised domain-independent aspect detection for sentiment analysis of customer reviews", Knowledge-Based Systems, 52, 2013; Genc-Nayebi et al., "A systematic literature review: Opinion mining studies from mobile app store user reviews”, Journal of Systems and Software, 125, 2017). On the other hand, it is known that log data from a system is used to detect a malicious attack on a system (Kim et al., "Long Short Term Memory Recurrent Neural Network Classifiers for Intrusion Detection", International Conference on Platform Technology and Service, 2015; Tuor et al., "Deep Learning for Unsupervised Insider Threat Detection in Structured Cybersecurity Data Streams, arXiv: 1710:00811v2, 2017) or to detect an error in data generated by the system or the system itself (Min et al., "DeepLog: Anomaly Detection and Diagnosis from System Logs through Deep Learning", CCS: Comuter and Communications Security, 2017. Zhang et al., "Automated IT system failure prediction: A deep learning approach", IEEE International Conference on Big Data, 2016) or to predict maintenance for a specific component of the system (Sipos et al., "Log-based predictive maintenance", 20 ^th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2014; US2015/0227838A1 ).

What all of these methods have in common is that only one source of information, for example customer service data (survey data or posts, etc.) or log data (log data, etc.) of the medical device, is used to determine customer satisfaction or information about the system of the medical device to determine the device.

It is therefore the object of the present invention to provide a method which makes it possible to determine the satisfaction of a customer based on log data and customer service data.

The object is achieved by a method for providing satisfaction information about a predicted satisfaction of a customer with regard to a medical device, by a method for providing a first trained function, by a system for providing satisfaction information for a customer with regard to a medical device, by a computer program product and by a computer-readable storage medium according to the independent claims. Advantageous developments are listed in the dependent claims and in the following description.

The solution to the problem according to the invention is described below both in relation to the claimed devices or systems and in relation to the claimed method. Features, advantages or alternative embodiments mentioned here are also to be transferred to the other claimed subjects and vice versa. In other words, the subject claims (which are directed, for example, to a device) can also be developed with the features that are described or claimed in connection with a method being. The corresponding functional features of the method are formed by corresponding physical modules.

In the following, the solution of the task according to the invention in relation to the claimed method and the claimed systems for providing satisfaction information about a predicted satisfaction of a customer with regard to a medical device as well as in relation to the claimed method and the claimed systems for training is also described a first trained function described. Features, advantages or alternative embodiments mentioned here are also to be transferred to the other claimed subjects and vice versa. In other words, the method and system claims for training the first trained function can also have features that are described or claimed in connection with the method and systems for providing satisfaction information about a customer's predicted satisfaction with regard to a medical device , be educated and vice versa.

In particular, the method and the system for providing the first trained function can be adapted to the method and the system for providing satisfaction information about a predicted satisfaction of a customer with regard to a medical device. In addition, input data of the method for providing satisfaction information can include advantageous features and implementations of the training input data, and vice versa. In addition, output data of the method for providing satisfaction information can include advantageous features and implementations of the training output data, and vice versa.

The invention relates to a computer-implemented method for providing satisfaction information about a customer's predicted satisfaction with regard to a medical device. The method includes the method step of providing input data, the input data including at least one operating parameter of the medical device and at least one item of customer information. The method also includes the step of applying a first trained function to the input data, thereby generating the satisfaction information. The method also includes the step of providing the satisfaction information.

The satisfaction information describes in particular the satisfaction of the customer for a user. In particular, the user can be part of a customer service. The customer service can in particular support or advise the medical device or the customer. In particular, the user can be a service technician or a service employee or a maintenance technician or a software technician or a customer support employee etc. In this case, the satisfaction information is predicted in particular for a period of time in the future. In other words, the satisfaction information includes the customer's predicted satisfaction. Customer satisfaction relates in particular to a medical device. In this case, the satisfaction can relate in particular to the way the medical device works and/or to the reliability of the medical device and/or to customer service with regard to the medical device and so on.

The medical device can in particular include a device for medical laboratory tests, e.g. a device for preparing or testing laboratory samples for in-vitro tests or a device for laboratory automation. The medical device can in particular include an imaging medical device. The medical imaging device can in particular be an X-ray device and/or a computed tomography (CT) device and/or a magnetic resonance tomography (MRT) device and/or a C-arm and/or a positron emission tomography ( PET) device and/or a single photon emission computed tomography (SPECT) device and/or an ultrasound imaging device. Alternatively, the medical device can include a patient couch and/or a robotic system for examination and/or operation support and/or a software system, etc. The software system can be designed in particular for displaying and/or analyzing and/or processing medical image data. In particular, the medical device can include any hardware or software in a medical or clinical context. The medical device can in particular also be a plurality of medical devices or a combination of medical devices of the type mentioned above. In this way, the invention can be used to predict or to provide satisfaction information about a predicted satisfaction of a customer with regard to a fleet of devices or a group of devices.

In the providing method step, in particular the input data are provided for further processing of the input data. The provision of the input data can go into include receiving the input data in particular. In this case, the input data can be provided in particular by the medical device. Alternatively or additionally, the input data can be provided by a customer service system on which customer information is recorded. Alternatively or additionally, the input data can be provided by a cloud system. Alternatively or additionally, the input data can be provided by an internal database. The input data include at least one operating parameter of the medical device and at least one piece of customer information.

The operating parameter describes in particular a mode of operation and/or a use and/or an environmental parameter or an environmental condition and/or a performance etc. of the medical device. In particular, the operating parameter describes a technical aspect of the medical device, in particular the operating parameter can include information regarding a type of use, duration of use, frequency of use, type of error, duration of error, frequency of errors, maintenance status and similar information. Alternatively or additionally, the operating parameter can include information about the frequency of an abort and/or a restart of a specific process, for example an examination. Alternatively or additionally, the operating parameter can include information about a system restart and/or a sub-system restart and/or their frequency. Alternatively or additionally, the operating parameter can include information about an external parameter of the medical device. The external parameter can in particular include a power supply or power grid stability of the medical device and/or a data network connection or data network stability of the medical device and/or an ambient temperature of the medical device etc. The operating parameter can in particular include a numerical value which describes the medical device or its function, etc. For example, such a numerical value can describe a number of failures of the medical device. Alternatively or additionally, the operating parameter may include an alphabetic value. For example, such an alphabetical value can describe whether a setting of the medical device is “on” or “off”. Alternatively or additionally, the operational parameter may include an alphanumeric value. In particular, the alphanumeric value can be a pair of values consisting of an alphabetic value and a numeric value. For example, the alphanumeric value may include a descriptive portion such as "ambient temperature in degrees Celsius" and a value such as "25" for that descriptive portion.

The customer information relates in particular to the customer whose satisfaction is to be predicted. In particular, the customer information includes at least one piece of information about the customer. The customer information describes in particular a behavior of the customer and/or a frequency of attempts by the customer to contact customer service and/or a number of medical devices that the customer owns and/or information about spare parts that the customer has already received or has ordered, etc. The customer information can in particular include a numeric value, an alphabetical value and/or an alphanumeric value.

In the method step of applying the first trained function, the satisfaction information is generated using the first trained function based on the input data.

In general, a trained function mimics cognitive functions that humans associate with human thinking. In particular, through training based on training data, the trained function can adapt to new circumstances and recognize and extrapolate patterns.

In general, parameters of a trained function can be adjusted using training. In particular, supervised training, semi-supervised training, unsupervised training, reinforcement learning and/or active learning can be used for this. In addition, representation learning (an alternative term is “feature learning”) (representation learning or feature learning) can be used. In particular, the parameters of the trained functions can be iteratively adapted through a number of training steps.

In particular, a trained function may include a neural network, a support vector machine, a decision tree, and/or a Bayesian network, and/or the trained function may be based on k-means clustering (k-means clustering), Q-learning, genetic algorithms and/or association rules. In particular, a trained function can comprise a combination of several uncorrelated decision trees or an ensemble of decision trees (random forest). In particular, the trained function can be determined using XGBoosting (extreme gradient boosting). In particular, a neural network werk may be a deep neural network, a convolutional neural network, or a convolutional deep neural network. In addition, a neural network may be an adversarial network, a deep adversarial network, and/or a generative adversarial network. In particular, a neural network can be a recurrent neural network. In particular, a recurrent neural network can be a long-short-term-memory (LSTM) network, in particular a gated recurrent unit (GRU). In particular, a trained function can include a combination of the described approaches. In particular, the approaches described here for a trained function are called the network architecture of the trained function.

In the step of providing the satisfaction information, the satisfaction information is provided to the user. The provision of the satisfaction information can in particular display the satisfaction information and/or send the satisfaction information, for example via e-mail or SMS and/or store the satisfaction information in a memory or an external database or a cloud -Memory include.

The inventors have recognized that it is possible to predict a customer's satisfaction. The inventors have recognized that such a prediction can be generated or determined in particular on the basis of at least one operating parameter and at least one piece of customer information. In other words, the inventors have recognized that based on a combination of operating data and customer information, customer satisfaction can be predicted and provided to the user as satisfaction information. In particular, the inventors have recognized that customer satisfaction can be predicted particularly reliably based on at least one piece of technical information about the medical device, the operating parameter, and customer data, the customer information.

According to one aspect of the invention, the method also includes the method step of determining the at least one operating parameter from log data of the medical device for a first defined period of time.

The log data can in particular include a log file or log file or log data. In particular, the log data can include an event log file or an event log file. In particular, processes that run in a computer system or computer system and/or a network of the medical device are logged in a log file. In particular, processes that run on the medical device are documented in the log file. In particular, the log file can include information regarding the use, functionality, stability, etc. of the medical device. Alternatively or additionally, the log data can include a stability parameter of the medical device. In particular, the stability parameter can include information about an abort of an image recording and/or about a pop-up and/or about a software update etc. Alternatively or additionally, the log data can include at least one environmental parameter or environmental condition. The environmental parameter can in particular be a temperature, humidity, a country in which the medical device is located, etc.

The first defined period of time can in particular lie in the past. In other words, the first defined period of time can be before a point in time at which the at least one operating parameter is determined. The first defined period of time defines a period of time for which the at least one operating parameter is determined from the log data. In particular, a time profile of the at least one operating parameter can be determined within the first defined period of time. The first defined period of time can in particular include one week or two weeks or three weeks or four weeks or one month or two months or three months or half a year, etc. In particular, the first defined period of time can be longer or shorter than the examples listed here. In particular, the first defined period of time can lie between two examples listed here. In this context, defined means that a length or duration of the first defined time period can be specified or defined. In particular, the duration of the first defined period of time can be predefined. Alternatively, the duration of the first defined period of time can be defined by the user. In other words, the user can specify or define for which period of time the at least one operating parameter is to be determined. In particular, the user can define the first defined period of time using calendar data. Alternatively, the user can define the duration in days or weeks or months. In particular, a start of the first defined time period can then be defined from the point of view of the day on which the user defines the duration of the first defined time period or on which the operating parameter is determined.

In the method step of determining the at least one operating parameter from the log data, data relevant in particular to the satisfaction information can be extracted from the log data as operating parameters. In particular, more than one operating parameter can be determined from the log data.

The inventors have recognized that information in the form of the at least one operating parameter can be determined from the log data, which can have an influence on the satisfaction information. In addition, the inventors have recognized that the technical aspect of the medical device is taken into account by the log data when determining the satisfaction information.

According to a further aspect of the invention, the method also includes the method step of determining the at least one item of customer information based on sales data and/or customer service data for a second defined period of time.

In particular, the sales data and/or customer service data can relate to the medical device. In other words, the sales data and/or the customer service data can indicate information that is directly or indirectly related to the medical device. Information that is directly related to the medical device directly indicates data that relate to the medical device. Information that is indirectly related to the medical device indicates data that is provided, for example, by the customer about the medical device. Alternatively, the sales data and/or the customer service data can be related to the customer. In particular, the sales data and/or the customer service data can include information about the customer.

In particular, the sales data can include information about spare parts that have already been delivered or installed for the medical device. In particular, the sales data can include information about ordered spare parts for the medical device. Alternatively or additionally, the sales data can include a number of medical devices that the customer owns or that are looked after, used, or managed by the customer. Alternatively or additionally, the sales data may include customer costs. In particular, the customer's costs in relation to the medical device can be specified. In other words, the costs can indicate how much the customer spent on the medical device and/or on maintenance work and/or on repair work and/or on replacement parts, etc. Alternatively or additionally, the costs can be specified as to how much the customer has already invested in medical devices that are managed by customer service.

In particular, the customer service data can include information about the customer. In particular, the information about the customer can include, for example, the location of the customer or the country in which the customer is located and/or a period of time for how long the customer has owned or operated or used a medical device supported by the customer service and /or which medical device the customer owns or operates or uses, etc. Alternatively or additionally, the customer service data can include information about one or more of the customer's tickets. In particular, the customer can create a ticket if he has a problem with or a question about the medical device. In particular, the ticket can be sent to customer service. In particular, the customer service data can include information about the number of tickets and/or the age of a ticket and/or the processing status of a ticket and/or the type or category of a ticket. In particular, the type or category of a ticket can describe whether it is a regional or a global ticket. Alternatively or additionally, the type or category of the ticket can describe where or by whom the ticket is processed. In particular, this can include information as to whether it is a technical ticket, a maintenance ticket, a repair ticket, a complaint ticket, a question ticket, etc. Alternatively or additionally, the type or category of the ticket can describe the escalation level in which the ticket is located. The escalation level can be determined by the customer or by the user. The escalation level can be specified on a discrete or continuous scale. A high value on the scale can indicate a well advanced escalation. Alternatively or additionally, the customer service data can include information on how often a maintenance or repair technician was at the customer's site.

The second defined period of time can in particular lie in the past. In other words, the second defined period of time can be before a point in time at which the at least one item of customer information is determined. The second defined period of time defines a period of time in which the at least one item of customer information is determined from the sales data and/or customer service data. In particular, a time profile of the at least one piece of customer information can be determined within the second defined period of time. The second defined time span can in particular include one week or two weeks or three weeks or four weeks or one month or two months or three months or half a year, etc. In particular, the second defined period of time can be longer or shorter than the examples listed here. In particular, the second defined period of time can lie between two examples listed here.

In particular, the second defined time span can be different from the first defined time span. In particular, the first and second defined periods of time can be of different lengths. In particular, the first and the second defined period of time can be shifted in time with respect to one another. In particular, the first and second defined periods of time can overlap in time. In particular, the first and second defined periods of time can be disjunctive to one another. In other words, the first and second defined periods of time cannot overlap.

Alternatively, the first and second defined periods of time can be identical. In other words, the first defined period of time can be equal to the second defined period of time.

In this context, defined means that a length or duration of the second defined time period can be predetermined or defined. In particular, the duration of the second defined period of time can be predefined. Alternatively, the duration of the second defined period of time can be defined by the user. In other words, the user can specify or define for which period of time the at least one operating parameter is to be determined. In particular, the user can define the second defined period of time using calendar data. Alternatively, the user can define the duration of the second defined period of time in days or weeks or months. In particular, a start of the second defined time period can then be defined from the point of view of the day on which the user defines the duration of the second defined time period or on which the operating parameter is determined.

In the method step of determining the at least one item of customer information, the information that is relevant for a prediction of customer satisfaction or for the customer information can be extracted from the sales data and/or the customer service data as customer information could be. In particular, in the method step of determining the at least one piece of customer information, more than one piece of customer information can be determined.

The inventors have recognized that the sales data and/or the customer service data include information which is relevant to the satisfaction information. In particular, the inventors have recognized that information from the sales data and/or customer service data can influence customer satisfaction.

According to a further aspect of the invention, the at least one operating parameter and/or the at least one item of customer information in the method steps of determining the at least one operating parameter or the at least one item of customer information can be based on data from a product life cycle management Management, PLM) and/or from a supply chain management (Supply-Chan-Management, SCM).

In particular, product life management data includes information that arises during a development process and during the entire life cycle of the medical device. In particular, supply chain management data includes all information about a supply chain of the medical device. The supply chain begins in particular with the manufacture of the medical device and ends with the installation of the medical device at the customer. The supply chain thus also includes, in particular, the transport of the medical device, such as shipping the medical device.

The inventors have recognized that the at least one operating parameter and/or the at least one piece of customer information can also be determined from data from product life cycle management and/or from supply chain management. These can then be used in particular as input data for the first trained function. The inventors have recognized that this data can in particular include information that can explain later failures or problems with the medical device and can predict satisfaction information. For example, problems during transport or production can favor the failure of certain parts of the medical device. This failure in turn has an impact on the satisfaction information regarding the customer.

According to a further aspect of the invention, the determination of the at least one operating parameter and/or the at least one piece of customer information can be carried out by a feature extraction algorithm. The feature extraction algorithm optionally includes a second trained function.

In this case, the determination of the at least one operating parameter, in particular by one first feature extraction algorithm are executed. In this case, the determination of the at least one piece of customer information can be carried out in particular by a second feature extraction algorithm.

The feature extraction algorithm can be designed in particular to extract features from the log data and/or sales data and/or customer service data that can influence the satisfaction information in the form of the operating parameter or the customer information. In particular, the at least one operating parameter and/or the at least one item of customer information can be determined by means of the feature extraction algorithm. In particular, the feature extraction algorithm can be adapted to what information is to be extracted or determined using the feature extraction algorithm.

In this case, the feature extraction algorithm can be created in particular by an expert. In particular, the expert can define rules according to which the feature extraction algorithm determines the at least one operating parameter and/or the at least one piece of customer information. Alternatively or additionally, the feature extraction algorithm can determine the at least one operating parameter and/or the at least one item of customer information by means of pattern recognition. Alternatively or additionally, the feature extraction algorithm may include a counting algorithm that counts a particular feature in the log data and/or the sales data and/or the customer service data. In this way, for example, the number of tickets for the customer can be determined using the counting algorithm based on the customer service data. Alternatively or additionally, a number of aborted scans can be determined from the log data in this way, for example.

In particular, the feature extraction algorithm may include a second trained function. For this purpose, the second trained function for determining the at least one operating parameter can be automatically trained on the log data, for example. In particular, the feature extraction algorithm for determining the at least one operating parameter includes a sequence recognition algorithm (sequence mining or sequence pattern mining). Patterns can be recognized in partially structured data. To determine the at least one item of customer information, the second trained function can in particular include “natural language processing” for analyzing text data or alphabetic data or alphanumeric data.

In particular, the feature extraction algorithm can include a combination of the functions or algorithms described.

In particular, the feature extraction algorithm can access the log data and/or the sales data and/or the customer service data using a Python API.

The feature extraction algorithm can in particular include pre-processing of data. The data can in particular include the log data and/or the sales data and/or the customer service data and/or the at least one operating parameter and/or the at least one item of customer information. By means of the pre-processing, the at least one operating parameter and/or the at least one piece of customer information is processed in particular in such a way that it is suitable as input data for the first trained function. In particular, the log data and/or the sales data and/or the customer service data can be processed by means of the preprocessing in such a way that the at least one operating parameter or the at least one piece of customer information can be determined therefrom.

The inventors have recognized that the at least one operating parameter and/or the at least one item of customer information can be determined automatically using the feature extraction algorithm. In particular, the inventors have recognized that the at least one operating parameter and/or the at least one piece of customer information can be preprocessed with the feature extraction algorithm in such a way that they are suitable as input data for the first trained function.

According to a further aspect of the invention, the first and/or second time span comprises a plurality of disjoint time blocks. The disjoint time blocks follow one another in terms of time. In this case, the at least one operating parameter or the at least one item of customer information is determined in an accumulated manner for each of the time blocks.

In particular, a time block can include a time subsection or a time interval or a time interval of the first and/or second defined time span. In particular, disjunctive means that the time blocks of a defined period of time do not superimpose or overlap. In particular, the disjunctive time blocks of a defined period of time can directly follow one another in terms of time. In other words, the time blocks of a defined period of time can be uninterrupted connect to each other. In particular, the time blocks of the plurality of disjoint time blocks can be of the same size or of the same length. In other words, the blocks of time may be of equal duration. Alternatively, the time blocks can be of different sizes or lengths. In particular, the first defined period of time can be divided into a first plurality of disjoint time blocks. In particular, the second defined period of time can be divided into a second plurality of disjoint time blocks. In particular, the first plurality of disjoint time blocks may correspond to the second plurality of disjoint time blocks. In particular, the number of disjoint time blocks for the first and/or second defined time period is specified by the length or duration of the corresponding first and/or second defined time period and/or by the length or duration of the time blocks. In particular, the first and/or second defined period of time can comprise a block of time. Alternatively, the first and/or second defined period of time can comprise more than one block of time.

In particular, a time block can include a week or a month, for example.

In particular, “accumulated” means that the at least one operating parameter or the at least one item of customer information includes data over the entire time block. In particular, this can mean that the data over the time block are averaged over time, or that the data over the time block are summed up, or that the data over the time block are collected, for example, in a list, etc. In other words, a chronological progression of the at least an operating parameter or the at least one piece of customer information is determined in time steps with the size or duration of a time block.

The inventors have recognized that fluctuations can be compensated for by accumulating the at least one operating parameter or the at least one item of customer information. For example, a fluctuation in the at least one operating parameter or the at least one item of customer information over the weekend can be compensated for with a time block that includes a week. The inventors have recognized that by dividing the first and/or second defined period of time into disjoint time blocks, a time profile of the at least one operating parameter or of the at least one customer parameter can be determined. The inventors have also recognized that the progression over time can serve as input data for the first trained function for determining the satisfaction information and that this leads to an improvement in the predicted customer satisfaction or the satisfaction information.

According to a further aspect of the invention, the satisfaction information is generated for at least one prediction time block. The at least one prediction time block follows the first and/or second defined period of time.

In particular, the forecast time block can include a day or a week or a month etc. In particular, the satisfaction information or the predicted satisfaction of the customer with regard to the medical device can be determined within the prediction time block. In particular, the prediction time block can be in the future at a point in time when the satisfaction information is determined.

In particular, the satisfaction information can be determined for a plurality of disjoint prediction time blocks. In particular, the disjoint prediction time blocks can follow one another in time. In particular, satisfaction information can be generated for each of the disjunctive time blocks. In particular, a time course of the customer's satisfaction can be predicted.

The inventors have recognized that providing the satisfaction information for at least one prediction time block gives the user a sense of how the customer's satisfaction is evolving over time. In this way, the user can also estimate how much time he has to react and to satisfy the customer.

According to a further aspect of the invention, the satisfaction information includes at least one piece of classification information.

In particular, the customer's satisfaction can be classified using the classification information. In particular, the classification information indicates a measure of customer satisfaction with regard to the medical device.

In particular, customer satisfaction can be classified into discrete classes. In other words, the classification information can indicate an assignment of the customer's satisfaction to one of the discrete classes. In other words, the classification information indicates which class the customer's satisfaction was or is assigned to. For example, customer satisfaction can be divided or assigned into four classes. An assignment in class "1" can mean that the customer is very satisfied and has no complaints. An assignment in class "4" can indicate a highest escalation level. In other words, a customer whose satisfaction information includes class “4” classification information is very dissatisfied. Alternative class divisions are possible. In particular, the classification can be in more or fewer classes than four. Alternatively, the largest class, for example class "4", can indicate that the customer is very satisfied and class "1" indicates the highest escalation level. In particular, the predicted satisfaction of the customer can be classified in the same way as school grades. In particular, the classification of the predicted satisfaction can be divided into two classes: "0" means that the customer is satisfied and "1" means that the customer is very dissatisfied. Alternatively, the meaning of "0" and "1" may be reversed. Alternatively, the discrete classes can be denoted by words rather than numbers. For example, customer satisfaction can be symbolically described using a temperature scale. For example, the classes can be named as follows: "cold", "lukewarm", "warm" and "hot". "Cold" means that the customer is satisfied and "hot" that the customer is very dissatisfied and the highest escalation level has been reached.

Alternatively, the classification information may include an indication of customer satisfaction along a continuous scale. In particular, the scale includes a plurality of continuous classes. In particular, the scale can include values between 1 and 10. In particular, the classification information can have any value between 1 and 10. In particular, the classification information can include the value between 1 and 10, which describes the satisfaction of the customer. In particular, a value of "1" can mean that the customer is very satisfied and a value of "10" that the customer is very dissatisfied. The values between 1 and 10 describe gradations of customer satisfaction between the two limit values. Alternatively, the meaning of "1" and "10" may be reversed. Alternatively, limit values other than 1 and 10 are also conceivable for the continuous scale. Alternatively, the limit values of the continuous scale can be "0" and "1". The customer's satisfaction is given as a probability of an escalation or of a high level of customer dissatisfaction.

The inventors have recognized that the customer's satisfaction can be shown to the user in a simple and clear manner by means of the classification information. In particular, the inventors have recognized that the user can easily derive actions from the classification information in order to improve or ensure customer satisfaction.

According to a further aspect of the invention, the satisfaction information item comprises at least one item of explanation information relating to the at least one item of classification information.

In particular, the explanatory information includes a reason or an explanation or an explanation as to why the customer's satisfaction was assigned to the class specified in the classification information. In particular, the explanatory information can indicate which of the input data (operating parameters and/or customer information) were decisive for the allocation to the class indicated in the classification information. In other words, the explanation information includes information about how the classification information came about. For example, the number of tickets in a certain period of time can be decisive for assigning customer satisfaction to a certain class.

If the first trained function comprises a random tree or a decision tree or an ensemble of decision trees (random forest) or an XGBoost, the explanation information can be determined by means of a tree explanation algorithm (tree explainer algorithm). If the first trained function includes a (deep) neural network, for example a recurrent neural network and/or a convolutional neural network and/or a long short-term memory and/or a gated recurrent unit, the explanatory information can be mechanism (sensitivity, attention mechanism) and/or by means of an approach to relevance propagation (relevance propagation approach) and/or by means of a deep explainer (deep explainer).

The inventors have recognized that the explanation information allows the user to understand why he received a specific piece of classification information for the customer. From this, the user can deduce whether the classification information makes sense and what actions he must or should carry out so that the customer's satisfaction can be improved and/or ensured.

According to a further aspect of the invention, the method further comprises the step of providing the satisfaction information in a decision support system and the step of deriving a hand Recommendation from the satisfaction information by the decision support system.

In particular, providing the satisfaction information can include displaying the satisfaction information by the decision support system (frequently: decision support system). In particular, the display can be graphical or pictorial and/or textual on an output medium. The output medium can in particular be a screen or a computer screen. In particular, the decision support system can include the output medium. In particular, the decision support system may include a graphical user interface (GUI). In particular, the satisfaction information can be provided using the GUI. Alternatively, the provision can also include sending a message, in particular an e-mail and/or an SMS (Short Message Service) to the user.

The recommendation for action can in particular describe which action or action or type of action the user should take or carry out or initiate in order to ensure or improve the customer's satisfaction or to prevent an escalation by the customer. The recommendation for action can be, for example, contacting the customer by telephone or e-mail and/or visiting the customer and/or making an offer for the customer (e.g. free use of a software add-on for a limited period of time, etc.) and/or or a prioritized treatment of the customer and/or a period in which the customer should be contacted at the latest etc. Alternatively or additionally, the recommendation for action can indicate which problem the customer has or whether it is a technical problem or a service problem. Alternatively or additionally, the recommendation for action can include a prioritization of a plurality of customers whose satisfaction information is provided. In other words, the recommended action can issue a recommendation as to which customer the user should focus on.

In particular, the recommendation for action can be derived from the decision support system based on the satisfaction information. In particular, the decision support system can derive an urgency of the recommended action based on the classification information. In particular, the decision support system can infer the type of action based on the explanation information. In particular, the recommended action can be provided in the decision support system. In particular, the recommended action can be displayed. In particular, the recommended action can be displayed together with the satisfaction information. In particular, the recommended action can be provided or displayed using the GUI of the decision support system.

In alternative embodiments, the user himself can derive the recommended action based on the satisfaction information.

The inventors have recognized that the provision of the satisfaction information allows the user to act in a targeted manner in order to ensure or ensure or improve the satisfaction of the customer. In addition, the inventors have recognized that the derivation of the recommended action by the decision support system supports the user in quickly deciding when which action is necessary or recommended in order to ensure customer satisfaction.

The invention further includes a computer-implemented method for providing a first trained function. The method includes the method step of providing training input data, the training input data including at least one operating parameter of a medical device and at least one piece of customer information. The method also includes the method step of providing initial training data, the initial training data including satisfaction information about a predicted satisfaction of the customer with regard to the medical device. The training output data and the training input data are related to one another. The method also includes the step of training the first trained function based on the training input data and the training output data. The method also includes the step of providing the first trained function.

In particular, the initial training data can be created by an expert or a user. In particular, the training input data and the training output data relate to a period in the past. In particular, the customer's satisfaction with regard to the training input data for creating the training output data is already known to the expert or user. In particular, the training output data can be derived by the expert or the user from the training input data. In particular, the training input data and the training output data in relation to each other.

In particular, in the method step of training the first trained function, the training can take place by means of supervised learning (supervised training) or unsupervised learning (unsupervised training). In particular, supervised learning may include random over-/under-sampling or synthetic minority over-sampling. In particular, an imbalance in the training input data and training output data in relation to classification information of the training output data can be compensated for. The classification information of the training output data is designed analogously to the classification information of the satisfaction information. In particular, the monitored learning can alternatively or additionally include cost-sensitive learning. In particular, an underrepresented class of the classification information of the initial training data can be weighted more heavily during the training due to the cost-sensitive learning. In particular, the unsupervised learning may include anomaly detection with a deep autoencoder model.

The inventors have recognized that historical data can be used as training input data. In particular, the inventors have recognized that the training baseline data for the past can be created by an expert or user and can be based on the customer's actual satisfaction in the past.

According to an optional aspect of the invention, the at least one operating parameter is determined for a first training period and the at least one item of customer information is determined for at least a second training period. The first training time span and the second training time span include a plurality of disjoint training time blocks. The training output data include the satisfaction information for at least one prediction training time block. The prediction-training time block follows the first and/or second training time period.

The first and the second training period of time can be configured analogously to the first and the second defined period of time. In particular, the disjoint training time blocks of the first and the second training time period can also be configured analogously to the disjoint time blocks of the first and the second defined time period. The prediction-training time block can be analogous to the prediction time block. However, the prediction training time block is in the past. In particular, the customer's satisfaction is known within the prediction training time block.

The inventors have recognized that an analog temporal description of the input data and output data for the training and the method for determining the satisfaction information enables training that is as efficient as possible and adapted to the data.

According to a further aspect of the invention, the training input data is recorded outside of an escalation time period. The escalation period is initiated by an escalation event.

In particular, the escalation event can be, for example, a complaint email from the customer to customer service and/or a call to customer service and/or a threat of consequences (e.g. contract termination) by the customer and/or a change of a customer to another provider etc. In particular, the escalation event or the start of the escalation period can be specified manually. Alternatively, the escalation event or the start of the escalation time period can be specified automatically. In particular, the escalation time period is then the time period or the time period or the time interval that is still influenced by the escalation event. In particular, the escalation event can be marked by an expert or a user. In particular, the escalation period of time directly follows the escalation event. In particular, the escalation time period can include one week or two weeks or three weeks or one month after the escalation event. In other words, the escalation period may be one week or two weeks or three weeks or one month after the escalation event. In particular, the duration can include a value between or greater than or less than the enumerated values. In particular, the duration of the escalation period can depend on the type of escalation event. For example, a customer's contract change can initiate a longer escalation period than a complaint email. In particular, the duration of the escalation period can be specified or determined by the expert or the user.

In particular, the training input data is recorded in such a way that it is not influenced by an escalation event, ie outside it the escalation period. In particular, the initial training data are also recorded or determined in such a way that they lie outside the escalation period.

The inventors have recognized that an escalation event can affect the training of the first trained function, so that it may not be possible for the first trained function to analyze the input data independently. The inventors have recognized that data which is outside of an escalation time period is preferably used for training the first trained function.

According to a further aspect of the invention, the first trained function trained according to the method described according to the invention can be used to provide the satisfaction information.

According to a further aspect of the invention, the first trained function is trained continuously by means of feedback. The feedback is based on a correspondence value between the provided satisfaction information and a determined satisfaction of the customer.

In particular, the user can subsequently determine or ascertain the ascertained or actual satisfaction of the customer. The ascertained satisfaction of the customer can be ascertained after the prediction time block for the satisfaction information has elapsed, for example based on an experience of the user or feedback from the customer. The user's experience may include, for example, a contract renewal or contract termination or an evaluation of the user.

In particular, the correspondence value can then be determined by a comparison between the provided satisfaction information of the predicted satisfaction and the ascertained satisfaction of the customer. In particular, the match value can be determined by the expert and/or the user. In particular, the match score may include a score on a continuous scale or a discrete class. For example, the match score may include grades analogous to school grades. In particular, the match score may include a class "1" when there is a very good match and a class "6" when there is no match.

The inventors have recognized that in this way the first trained function can be continuously improved and adapted. The inventors have also recognized that this can also improve the assessment of the customer's predicted satisfaction and the objectivity of this assessment.

According to a further aspect of the invention, the first trained function is selected from a plurality of first trained functions. The selection is based on the match score.

In particular, this method is known under the term "model selection".

In particular, the selection can take place during the training of the first trained function. In particular, a plurality of first trained functions can be trained during the training. In particular, the first trained functions can be different with regard to their mode of operation or network architecture. Exemplary network architectures are described above. In particular, the match value can be determined during the training. In this case, the agreement value is determined on the basis of the initial training data and the satisfaction information predicted by the first trained function. In other words, the initial training data is compared with the satisfaction information determined by the first function during the training. From this comparison, the match value can be determined. The correspondence value can be designed as described above. In particular, the match value can be determined automatically or manually. In particular, that first trained function can then be selected whose ascertained satisfaction information has the best match value with the initial training data.

In particular, the first trained function can alternatively or additionally be selected during the execution of the method according to the invention. In particular, the satisfaction information can be determined in parallel for each of the first trained functions. In this case, the satisfaction information is only provided to the user by the selected first trained function. Through the feedback from the user, the match value can be determined as described above for each of the first trained functions. Based on the match value, the satisfaction information that was determined by the first trained function with the best match value can be provided the next time or when the method is carried out repeatedly. In other words, the selected first trained function can be replaced by another first trained function if, according to the feedback, their match calculation value is better than that of the originally selected first trained function. In this case, the selection can alternatively be based on an average of a number of match values for a number of items of satisfaction information. In particular, it can be checked continuously which first trained function is best suited or has the best match value.

The inventors have recognized that the most suitable first trained function for predicting the customer's satisfaction or for determining the satisfaction information can be flexibly selected by means of "model selection". In this way, it can be ensured that the satisfaction information that is most suitable in terms of the match value is provided.

The invention also includes a system for providing satisfaction information about a customer's predicted satisfaction with regard to a medical device. The system includes a computing unit and an interface. In this case, the arithmetic unit is designed to provide input data. The input data include at least one operating parameter of the medical device and at least one piece of customer information. In this case, the arithmetic unit is also designed to apply a first trained function, as a result of which the satisfaction information is generated. The interface is designed to provide the satisfaction information.

Such a system can in particular be designed to carry out the previously described method for providing satisfaction information about a predicted satisfaction of a customer with regard to a medical device and its aspects. The system is designed to carry out this method and its aspects in that the interface and the computing unit are designed to carry out the corresponding method steps.

The invention also relates to a computer program product with a computer program and a computer-readable medium. A largely software-based implementation has the advantage that systems already in use can be retrofitted in a simple manner by means of a software update in order to work in the manner described. In addition to the computer program, such a computer program product may include additional components such as documentation and/or additional components, as well as hardware components such as hardware keys (dongles, etc.) for using the software.

In particular, the invention also relates to a computer program product with a computer program that can be loaded directly into a memory of a system, with program sections for all method steps of the method described above for providing satisfaction information about a predicted satisfaction of a customer with regard to a medical device and its aspects to be executed when the program sections are executed by the system.

In particular, the invention relates to a computer-readable storage medium on which program sections that can be read and executed by a system are stored in order to carry out all the method steps of the method described above for providing satisfaction information about a predicted satisfaction of a customer with regard to a medical device and its aspects if the Program sections are executed by the system.

The invention also relates to a training system for providing a first trained function. The training system includes a training interface and a training processor. The training processing unit is designed to provide training input data. The training input data include at least one operating parameter of a medical device and at least one piece of customer information. In this case, the training processor is also designed to provide initial training data. In this case, the initial training data include satisfaction information about a predicted satisfaction of the customer with regard to the medical device. The training output data and the training input data are related to one another. The training processing unit is also designed to train the first trained function based on the training input data and the training output data. The training interface is designed to provide the first trained function.

The invention also relates to a training computer program product with a training computer program and a computer-readable training medium. A largely software-based realization has the advantage that even previously used training systems can be retrofitted in a simple manner by means of a software update in order to work in the manner described. Such a training computer program product can, in addition to the training computer program, optionally include additional components such as B. documentation and / or additional components, as well as hardware components, such as include hardware keys (dongles, etc.) to use the software.

In particular, the invention also relates to a training computer program product with a training computer program, which can be loaded directly into a memory of a training system, with program sections to carry out all method steps of the method described above for providing a first trained function and its aspects if the Program sections are executed by the training system.

In particular, the invention relates to a computer-readable training storage medium on which program sections that can be read and executed by a training system are stored in order to carry out all the method steps of the method described above for providing a first trained function and its aspects if the program sections are stored by the training system to be executed.

The characteristics, features and advantages of this invention described above will become clearer and more understandable in connection with the following figures and their descriptions. The figures and descriptions are not intended to restrict the invention and its embodiments in any way.

The same components are provided with corresponding reference symbols in different figures. The figures are generally not to scale.

• 1 ein erstes Ausführungsbeispiel eines Verfahrens zum Bereitstellen einer Zufriedenheits-Information über eine vorhergesagte Zufriedenheit eines Kunden bezüglich eines medizintechnischen Gerätes,
• 2 ein zweites Ausführungsbeispiel eines Verfahrens zum Bereitstellen einer Zufriedenheits-Information über eine vorhergesagte Zufriedenheit eines Kunden bezüglich eines medizintechnischen Gerätes,
• 3 ein drittes Ausführungsbeispiel eines Verfahrens zum Bereitstellen einer Zufriedenheits-Information über eine vorhergesagte Zufriedenheit eines Kunden bezüglich eines medizintechnischen Gerätes,
• 4 ein Ausführungsbeispiel von einer definierten Zeitspanne umfassend eine Mehrzahl an disjunkten Zeitblöcken und einem Vorhersage-Zeitblock,
• 5 ein Ausführungsbeispiel eines Verfahrens zum Bereitstellen einer ersten trainierten Funktion,
• 6 ein Ausführungsbeispiel einer Trainings-Zeitspanne umfassend eine Mehrzahl an disjunkten Trainings-Zeitblöcken, eine Vorhersage-Trainings-Zeitspanne, eine Eskalations-Zeitspanne und ein Eskalations-Ereignis,
• 7 ein System zum Bereitstellen einer Zufriedenheits-Information über eine vorhergesagte Zufriedenheit eines Kunden bezüglich eines medizintechnischen Gerätes, 8 ein Trainings-System zum Bereitstellen einer ersten trainierten Funktion.

Show it

• 1 a first exemplary embodiment of a method for providing satisfaction information about a predicted satisfaction of a customer with regard to a medical device,
• 2 a second exemplary embodiment of a method for providing satisfaction information about a predicted satisfaction of a customer with regard to a medical device,
• 3 a third exemplary embodiment of a method for providing satisfaction information about a predicted satisfaction of a customer with regard to a medical device,
• 4 an embodiment of a defined period of time comprising a plurality of disjoint time blocks and a prediction time block,
• 5 an embodiment of a method for providing a first trained function,
• 6 an embodiment of a training period comprising a plurality of disjoint training time blocks, a prediction training period, an escalation period and an escalation event,
• 7 a system for providing satisfaction information about a customer's predicted satisfaction with regard to a medical device, 8th a training system for providing a first trained function.

1 shows a first exemplary embodiment of a method for providing satisfaction information about a predicted satisfaction of a customer with regard to a medical device.

In the method step of providing input data PROV-01, input data for determining the satisfaction information is received from a system SYS for providing the satisfaction information. The data can be sent from the medical device and/or from a customer service system to the SYS system. Alternatively, the input data can be retrieved from the system SYS in the procedural step of providing PROV-01 the input data. In particular, the input data can be retrieved or made available from an internal database of the system SYS and/or from an external database. In particular, the external database can be stored on a cloud system and/or a server system. Alternatively, in the procedural step of providing the input data PROV-01, the input data can be determined or recorded in particular by the system SYS.

The input data include at least one operating parameter of the medical device and at least one piece of customer information.

The medical device can in particular include a device for medical laboratory tests, for example a device for preparing or testing laboratory samples for in-vitro tests or a device for laboratory automation. The medical device can in particular include an imaging medical device. The medical device can in particular be an X-ray device or a computed tomography (CT) device or a magnetic resonance tomography (MRT) device or a C-arm or a positron emission tomography (PET) device or a single photon Emission computed tomography (single photon emission computed tomography, SPECT) device and/or a device for ultrasound imaging. Alternatively, the medical device can include a patient couch or a robotic system or a software system. The software system can be designed in particular for displaying and/or analyzing and/or processing medical image data. In particular, the medical device can include any hardware or software in a medical or clinical context. The medical device can in particular also be a plurality of medical devices or a combination of medical devices of the type mentioned above. In this way, the invention can be used to predict or to provide satisfaction information about a predicted satisfaction of a customer with regard to a fleet of devices or a group of devices.

The operating parameter describes, for example, a use and/or an environmental parameter or an environmental condition and/or a performance and/or a mode of operation of the medical device. The use can describe, for example, which program or which function of the medical device is used and how often. The use can also describe a utilization of the medical device. The performance can indicate a measure of the effectiveness of the medical device. The effectiveness can in particular describe a length of time that the medical device needs to run a program. The mode of operation can in particular describe whether all components of the medical device are functioning as intended. The environmental parameter can in particular include a room temperature and/or a device temperature and/or a humidity and/or a country in which the medical device is located, etc. In particular, the operating parameter can include information relating to a type of use, a period of use, a frequency of use, a type of error, an error duration, an error frequency, a maintenance status, and similar information. Alternatively or additionally, the operating parameter can include information about the frequency of an abort and/or a restart of a specific process, for example an examination. Alternatively or additionally, the operating parameter can include information about a system restart and/or a sub-system restart and/or their frequency. Alternatively or additionally, the operating parameter can include information about an external parameter of the medical device. The external parameter can in particular include a power supply or power grid stability of the medical device and/or a data network connection or data network stability of the medical device etc. The operational parameter may include a numeric value, or an alphanumeric value, or an alphabetic value.

The customer information can in particular include a behavior of the customer and/or a frequency or number of contact attempts by the customer with a customer service and/or a number of medical devices that the customer owns or looks after. In other words, the customer information can include any information about the customer whose satisfaction information is to be provided. The customer information may include a numeric value, or an alphanumeric value, or an alphabetic value.

In the method step of applying APP the first trained function, the satisfaction information is determined or generated from the input data. The satisfaction information describes the predicted satisfaction of the customer for a period of time, a so-called prediction time block VZB in the future. The satisfaction information is based on the at least one operating parameter and the at least one piece of customer information. The satisfaction information includes at least one piece of classification information. The classification information describes the predicted satisfaction of the customer based on a discrete or continuous scale or classification. For example, the classification information “1” can be assigned to a customer who is predicted to be very satisfied. The classification information “4” can be assigned to a customer who is predicted to be very dissatisfied. Alternatively, for example, "4" can stand for very satisfied and "1" for very dissatisfied. The gradations between these classes can be discrete or continuous. The classes can be named alternatively, for example "satisfied" to "very dissatisfied". The classes can alternatively be divided according to the principle of school grades. The satisfaction information can also include explanatory information on the classification information. The explanation information describes how the classification information came about. In other words, the explanation information gives a reason why the customer's satisfaction was predicted according to the classification information. The explanation information indicates which of the input data was decisive for the corresponding class division of the classification information.

In the procedural step of providing the satisfaction information PROV-02, the satisfaction information generated in the procedural step of applying APP the first trained function is provided to a user. In particular, the user can be a customer service employee. The customer service can in particular be entrusted with the maintenance of the medical device or with the organization of the maintenance of the medical device and/or with the support of the customer. The satisfaction information can be provided by displaying the satisfaction information on a display medium or output medium, for example a screen. Alternatively, the satisfaction information can be provided in this method step by sending the satisfaction information to the customer, for example by SMS or e-mail.

2 shows a second exemplary embodiment of a method for providing satisfaction information about a predicted satisfaction of a customer with regard to a medical device.

The procedural steps of providing PROV-01 the input data, applying APP the first trained function and providing PROV-02 the satisfaction information become according to the description 1 executed.

In the method step of determining the at least one operating parameter DET-01, the at least one operating parameter is determined from log data of the medical device for a first defined period of time ZS.

The protocol data can in particular include at least one log file or one log file and/or one event log file of the medical device. The log data can, for example, include information that describes how (which function, how often, how long) the medical device is used, whether all components of the medical device work as intended, which parameters (e.g. movement parameters of a patient bed and/or a robot arm, Exposure time, X-ray voltage, etc.) are set. Alternatively or additionally, the log data can include information about an environmental parameter or an environmental condition of the medical device. The environmental parameter can be detected, for example, with a sensor of the medical device.

The first defined period of time ZS includes a period of time for which the at least one operating parameter is determined from the log data. The first defined period of time ZS can in particular include a plurality of disjunctive time blocks ZB01, ZB02, ZB03, ZB04, ZB05. The disjoint time blocks ZB01, . . . , ZB05 can, in particular, divide the first defined time span ZS into a number of intervals or time intervals. The disjoint time blocks ZB01, . . . , ZB05 can follow one another in terms of time without overlapping or superimposing one another. In particular, the time blocks ZB01, . . . , ZB05 can all be of the same size. The at least one operating parameter can be determined accumulated in particular for each time block ZB01, . . . , ZB05. In other words, the operating parameter can be determined individually for each time block ZB01, . . . , ZB05. In this case, for example, a number or frequency with which a program or a function of the medical device is executed can be determined summed up for a time block ZB01, . . . , ZB05. Alternatively, a mean value of the operating parameter can be determined over the corresponding time block ZB01, ..., ZB05 or a list of the operating parameters for the corresponding time block ZB01, ..., ZB05. Whether the sum, the mean value or a list of the operating parameters is determined for the corresponding time block ZB01, . . . ZB05 depends on the type of operating parameter or on what the operating parameter describes. A time block ZB01, ..., ZB05 can, for example, include a week or seven days. For example, fluctuations in the operating parameter can be compensated for over a weekend, since averaged or totaled or listed values for the at least one operating parameter are determined over a week. Alternatively, a time block ZB01, . . . , ZB05 can include one month, for example. The first defined period of time ZS can include any number of time blocks ZB01, . . . , ZB05. In particular, the first defined period of time ZS can include a time block ZB01, . . . , ZB05. In particular, the first defined period of time ZS can include more than one time block ZB01, . . . , ZB05. The first defined period of time ZS can in particular be specified or defined by a user. For this purpose, the user can specify, for example using calendar data, from when to when the first defined time span ZS should extend. Alternatively, the user can specify a duration that the first defined time period ZS should include. The first defined period of time ZS can end on the day on which the satisfaction information is generated. The first defined time period ZS then begins on the day that is determined starting from the end day according to the duration of the first defined time period ZS. Alternatively, the duration of the first defined time period ZS can be specified. In particular, the prediction time block VZB can directly follow the first defined time period ZS in terms of time.

In the method step of determining DET-02 the at least one piece of customer information, the at least one piece of customer information is determined based on sales data and/or customer service data for a second defined period of time ZS.

The sales data can in particular include information about the number and/or type of spare parts delivered and/or ordered for the medical device. Alternatively or additionally, the sales data can include information on how many medical devices the customer owns or looks after. Alternatively or additionally, the sales data can include costs that the customer already had in relation to the medical device.

The customer service data can in particular include information about the customer. This information can be derived from customer surveys, for example. Alternatively or additionally, the customer service data can contain information about a number and/or urgency and/or type of tickets that the customer has sent to the customer service. The type of ticket can describe where the ticket has to be processed, whether it is an inquiry, a complaint, a defect in the medical device, etc. In particular, tickets relating to the medical device for which the satisfaction -Information should be created, stand, be taken into account. Alternatively or additionally, all of the customer's tickets can be taken into account. In particular, open and already completed tickets can be taken into account. Alternatively or additionally, customer service data can also be derived from a conversation with the customer. Alternatively or additionally, customer service data can include information on how often a technician has already been to the customer's site. Alternatively or additionally, the customer service data can include information about a contract volume and/or a contract duration and/or other contract details of the customer.

The second defined time span ZS is analogous to the first defined time span ZS. In particular, the second defined time span ZS can correspond to the first defined time span US. In particular, the at least one item of customer information for a time block ZB01, . . . , ZB05 is determined in an accumulated manner. In particular, the customer information about the time block ZB01, ..., ZB05 can be averaged or summed up or listed.

Determining DET-01 the at least one operating parameter and/or determining DET-02 the at least one item of customer information may be performed using a feature extraction algorithm. The feature extraction algorithm can be created by an expert, for example. Alternatively or additionally, the feature extraction algorithm can determine the at least one operating parameter and/or the at least one item of customer information by means of analytical analysis. Alternatively or additionally, the feature extraction algorithm may include a second trained function. In implementations, the feature extraction algorithm may be part of the first trained function. In particular, the at least one operating parameter can be determined by a first feature extraction algorithm. In particular, the at least one item of customer information can be determined by a second feature extraction algorithm. In particular, the first and the second feature extraction algorithm can be different. In particular, the first feature extraction algorithm can include a sequence recognition algorithm (sequence mining or sequence pattern mining). In particular, the second feature extraction algorithm can include “natural language processing”.

3 shows a third exemplary embodiment of a method for providing satisfaction information about a predicted satisfaction of a customer with regard to a medical device.

The procedural steps of providing PROV-01 the input data, applying APP the first trained function and providing PROV-02 the satisfaction information become according to the description 1 executed. The method steps of determining DET-01 the at least one operating parameter and determining DET-02 the at least one piece of customer information become according to the description 2 executed.

In the step of providing the satisfaction information PROV-03, the satisfaction information is provided in a decision support system. In this case, the satisfaction information is made available to the user in the decision support system. The provision can take place in particular by displaying the satisfaction information through a display medium or output medium. The output medium can in particular be a screen or a computer screen. The satisfaction information can be displayed or provided in the decision support system in the form of images or graphics and/or text. In particular, the decision support system can include a graphical user interface (GUI). sen, by means of which the satisfaction information is presented or displayed.

In the method step DET-03 of deriving a recommendation for action, the recommendation for action is derived from the satisfaction information by the decision support system. The recommendation for action can indicate what action the user should take to improve or ensure customer satisfaction or to prevent escalation by the customer. The recommendation for action can be, for example, a recommendation for a phone call, sending spare parts, a customer visit, an offer to the customer or a discount campaign for the customer, waiting, answering a customer inquiry, etc. Alternatively or additionally, the recommendation for action can include a prioritization of a plurality of customers. According to the method according to the invention, satisfaction information was provided in advance for each customer of the plurality of customers. Depending on this satisfaction information from the majority of customers, the prioritization can be carried out. The prioritization indicates which customer is preferred or treated particularly quickly or which customer request should be processed particularly promptly.

The recommended action can be provided in the decision support system. The recommended action can be displayed using the display medium of the decision support system. The recommendation for action can be presented graphically or pictorially and/or in text.

4 shows an embodiment of a defined period of time ZS comprising a plurality of disjunctive time blocks ZB01, ..., ZB05 and a prediction time block VZB.

In the representation of 4 a time course is illustrated by the horizontal arrow. "t" stands for time. The first and second periods of time ZS mentioned in the description can both be designed according to the defined period of time ZS described here. The defined period of time ZS is divided into five disjoint time blocks ZB01, . . . , ZB05. The time blocks ZB01, . . . , ZB05 follow one another in terms of time. The time blocks ZB01, ..., ZB05 describe the entire defined period of time ZS. The at least one operating parameter and/or the at least one piece of customer information can be determined accumulated for each time block ZB01, . . . , ZB05 as described above. A time profile of the at least one operating parameter and/or the at least one item of customer information can thus be determined over the defined time period ZS. This time course can serve as input data for the first trained function.

The prediction time block VZB follows directly on the defined period of time. In the step of applying APP the first trained function, the satisfaction information for the prediction time block VZB is determined based on the at least one operating parameter and the at least one piece of customer information.

In particular, the defined period of time ZS in the past and the prediction time block VZB in the future. In other words, based on known data (operating parameters, customer information), the customer's satisfaction can be predicted using the satisfaction information.

5 shows an embodiment of a method for providing a first trained function.

In the method step of providing TPROV-10 training input data, the training input data are made available to a training system. The training input data include at least one operating parameter of a medical device and at least one piece of customer information. Providing TPROV-01 of the training input data can be analogous to providing PROV-01 of the input data, as in the description of 1 described, take place.

In the method step of providing TPROV-02 of initial training data, the initial training data is provided to the training system. The initial training data include satisfaction information about a predicted satisfaction of the customer with regard to a medical device. The training output data is related to the training input data. For this purpose, the training output data can have been determined by an expert or a user based on the training input data. In particular, the initial training data can have been created on the basis of observations or experiences of the expert or user in relation to the initial training data. In particular, feedback from the customer can be taken into account when creating it. Providing TPROV-02 of the training output data can take place analogously to providing PROV-01 of the input data.

In the method step of training TRAIN the first trained function, the first trained function is based on the training input data and the training output data trai ned. In particular, the first trained function is trained for this in such a way that satisfaction information generated by the first trained function and based on the training input data deviates as little as possible from the associated training output data. This deviation is quantified by a match value.

The method step of training TRAIN the first trained function can be carried out for a plurality of first trained functions.

In the step of providing TPROV-03 the first trained function, the first trained function is provided to the user so that he can use the first trained function when executing the method according to the invention for determining the satisfaction information. Here, when a plurality of first trained functions have been trained in the step of training TRAIN, the first trained function whose matching value is best can be provided.

While the method according to the invention is being carried out, the first trained function or the plurality of first trained functions can be trained further by means of feedback. For this purpose, further training output data for the input data are generated afterwards based on an experience or observation of the user. The further initial training data correspond to a level of satisfaction ascertained by the user. Alternatively, the user can specify the correspondence value between the predicted satisfaction of the customer in the satisfaction information and a determined or observed satisfaction of the customer. Based on this match value, the first trained function or the plurality of first trained functions can be trained continuously while the method according to the invention is being carried out. The input data serve as training input data. In particular, based on the match value of this training, another first function from the plurality of first trained functions can be provided if this proves to be more suitable based on the match value.

6 shows an embodiment of a training period TZS comprising a plurality of disjoint training time blocks TZB01, TZB02, TZB03, TZB04, TZB05 and a prediction training period VTZB, an escalation period EZS and an escalation event EE.

The training period TZS can be analogous to that according to 4 be formed described defined period of time ZS. The disjoint training time blocks TZB01, ..., TZB05 can analogous to the disjoint time blocks ZB01, ..., ZB05 according to 4 be trained. The prediction-training time block VTZB can be analogous to the prediction time block VZB 4 be trained. However, for the training, both the training time span TZB and the prediction training time block VTZB are in the past, so that the initial training data can be determined. The training input data are provided for the training period TZS. In this case, the training input data can be provided analogously to the input data for the plurality of disjoint training time blocks TZB01, . . . TZB05. The training output data are provided for the prediction training time block VTZB.

The representation also shows an escalation event EE and an escalation period of time ECB initiated by the escalation event EE. The escalation event EE can be an event initiated by the customer, which represents a high level of customer dissatisfaction. The escalation event can be, for example, a detailed complaint from the customer or a contract termination or termination, etc. The escalation period EZS is the period of time in which the customer's behavior and satisfaction is influenced by the escalation event EE. The duration of the escalation period EZS can depend on the escalation event EE.

The training time period TZS and the prediction training time block TZB lie outside the escalation time period EZS. In this way, a falsification of the training as a result of the escalation event EE can be avoided. The escalation event EE and the escalation period EZS can be determined and specified by an expert or a user.

In order to generate as much training input data and training output data as possible, the training period TZS and the prediction training period VTZS can be shifted along the time axis. Training input data and training output data can be generated for different positions. The training period TZS and the prediction training period VTZS are outside of the escalation period EZS. This shifting of the period of time can be done by a sliding window method.

7 shows a system SYS for providing satisfaction information about a customer's predicted satisfaction a medical device, 8th shows a training system TSYS for providing a first trained function.

The system SYS shown for providing the satisfaction information is designed to carry out a method according to the invention for providing the satisfaction information about the predicted satisfaction of the customer with regard to the medical device. The training system TSYS shown is designed to carry out a method according to the invention for providing the first trained function. The system SYS includes an interface SYS.IF, a processing unit SYS.CU and a memory unit SYS.MU. The training system TSYS comprises a training interface TSYS.IF, a training arithmetic unit TSYS.CU and a training memory unit TSYS.MU.

The system SYS and/or the training system TSYS can in particular be a computer, a microcontroller or an integrated circuit (IC). Alternatively, the system SYS and/or the training system TSYS can be a real or virtual computer network (a technical term for a real computer network is "cluster", a technical term for a virtual computer network is "cloud") . The system SYS and/or the training system TSYS can be designed as a virtual system which is executed on a computer or a real computer network or a virtual computer network (a technical term is “virtualization”).

The interface SYS.IF and/or the training interface TSYS.IF can be a hardware or software interface (e.g. a PCI bus, USB or Firewire). The processing unit SYS.CU and/or the training processing unit TSYS.CU can include hardware and/or software components, for example a microprocessor or a so-called FPGA (Field Programmable Gate Way). The storage unit SYS.MU and/or the training storage unit TSYS.MU can be used as non-permanent working memory (Random Access Memory, RAM) or as permanent mass storage (hard drive, USB stick, SD card, Solid State Disk (SSD)) be trained.

The interface SYS.IF and/or the training interface TSYS.IF can in particular include a number of sub-interfaces which execute different method steps of the respective method according to the invention. In other words, the interface SYS.IF and/or the training interface TSYS.IF can be designed as a plurality of interfaces SYS.IF and/or training interfaces TSYS.IF. The arithmetic unit SYS.CU and/or the training arithmetic unit TSYS.CU can in particular comprise a plurality of sub-arithmetic units which execute different method steps of the respective method according to the invention. In other words, the processing unit SYS.CU and/or the training processing unit TSYS.CU can be designed as a plurality of processing units SYS.CU and/or training processing units TSYS.CU.

Where this has not yet been done explicitly, but makes sense and is within the meaning of the invention, individual exemplary embodiments, their individual aspects or features can be combined with one another or exchanged without departing from the scope of the present invention. Advantages of the invention described with reference to one exemplary embodiment also apply to other exemplary embodiments, where applicable, without explicit mention.

QUOTES INCLUDED IN DESCRIPTION

This list of documents cited by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Patent Literature Cited

• US 2015/0227838 A1 [0003]

Claims (16)

Computer-implemented method for providing satisfaction information about a customer's predicted satisfaction with regard to a medical device, comprising - Providing (PROV-01) input data, the input data comprising at least one operating parameter of the medical device and at least one piece of customer information, - applying (APP) a first trained function to the input data, generating the satisfaction information, - Providing (PROV-02) the satisfaction information. procedure one claim 1 , further comprising the following method step: - determining (DET-01) the at least one operating parameter from log data of the medical device for a first defined period of time (ZS). Method according to one of the preceding claims, further comprising the following method step: - Determine (DET-02) the at least one customer information based on sales data and / or customer service data for a second defined period of time (ZS). Procedure according to one of claims 2 or 3 , wherein the determining of the at least one operating parameter and/or the at least one piece of customer information is performed by a feature extraction algorithm, wherein the feature extraction algorithm optionally comprises a second trained function. Procedure according to one of claims 2 until 4 , wherein the first and/or second defined period of time (ZS) comprises a plurality of disjunctive time blocks (ZB01, ZB02, ZB03, ZB04, ZB05), wherein the disjoint time blocks (ZB01, ..., ZB05) follow one another in terms of time, wherein the at least one operating parameter or the at least one item of customer information for each of the disjunctive time blocks (ZB01, . . . , ZB05) is determined in an accumulated manner. Procedure according to one of claims 2 until 5 , wherein the satisfaction information is generated for at least one prediction time block (VZB), the at least one prediction time block (VZB) following the first and/or second defined time period (ZS). Method according to one of the preceding claims, wherein the satisfaction information comprises at least one classification information. procedure after claim 7 , wherein the satisfaction information comprises at least one explanatory information on the at least one classification information. Method according to one of the preceding claims, further comprising the following method step: - Providing (PROV-03) the satisfaction information in a decision support system, - Deriving (DET-03) a recommendation for action from the satisfaction information by the decision support system. A computer-implemented method for providing a first trained function, comprising: - Providing (TPROV-01) training input data, the training input data comprising at least one operating parameter of a medical device and at least one item of customer information, - Providing (TPROV-02) training output data, the training output data including satisfaction information about a predicted satisfaction of the customer with regard to the medical device, the training output data and the training input data being related to one another, - train (TRAIN) the first trained function based on the training input data and the training output data, - Provide (TPROV-03) the first trained function. procedure after claim 10 , the training input data being recorded outside of an escalation period (EZS), the escalation period (EZS) being initiated by an escalation event (EE). Method according to one of the preceding claims, the first trained function is continuously trained by means of feedback, wherein the feedback is based on a match score between the provided satisfaction information and a determined customer satisfaction. procedure after claim 12 , wherein the first trained function is selected from a plurality of first trained functions, wherein the selecting is based on the match score. System (SYS) for providing satisfaction information about a customer's predicted satisfaction with regard to a medical device, comprising a processing unit (SYS.CU) and an interface (SYS.IF), the processing unit (SYS.CU) for providing (PROV -01) is formed from input data, wherein the input data include at least one operating parameter of the medical device and at least one piece of customer information, wherein the computing unit (SYS.CU) is also designed to apply (APP) a first trained function to the input data, whereby the satisfaction information is generated, wherein the interface (SYS.IF) is designed to provide (PROV-02) the satisfaction information. Computer program product with a computer program, which can be loaded directly into a memory (SYS.MU) of a system (SYS), with program sections for all method steps of the method according to one of Claims 1 until 9 to be executed when the program sections are executed by the system (SYS). Computer-readable storage medium on which a system (SYS) readable and executable program sections are stored to all method steps of the method according to one of Claims 1 until 9 to be executed when the program sections are executed by the system (SYS).

Images