CN118520303A - Automatic machine learning method and system based on structured data - Google Patents

Abstract

The present application discloses an automatic machine learning method and system based on structured data, which comprises the following steps: obtaining a structured data set and labeling target variables for the structured data; performing type identification on the structured data set to obtain a target task type and a target feature type; constructing a corresponding feature engineering according to the target feature type to obtain a target feature combination; constructing a base model according to the target task type and selecting a corresponding model fusion strategy; inputting the target feature combination and the target variable into the base model for training to obtain a target base model; fusing the target base model according to the model fusion strategy to obtain a fusion model, and automatically deploying the target base model and the fusion model; the present method greatly reduces the workload of manually selecting a model, and improves the versatility and flexibility of the model in various tasks, and can handle multiple tasks at one time, thereby saving a lot of time and energy and improving efficiency.

Description

An automatic machine learning method and system based on structured data

Technical Field

The present application relates to the technical field of machine learning, and in particular to an automatic machine learning method and system based on structured data.

Background Art

In today's information society, data has become the core driving force for technological progress and business development; structured data, such as tabular data in relational databases, has a wide range of applications in the field of machine learning due to its inherent regularity and predictability; however, traditional machine learning frameworks often require users to have a deep background in data science and programming in order to perform a series of tedious operations such as feature engineering, model selection, and parameter tuning; this not only increases the cost of learning, but also limits the application of machine learning technology in a wider range of fields. With the development of automation and intelligent technology, automatic machine learning (AutoML) has gradually become a research hotspot. AutoML aims to reduce manual intervention in the machine learning process through automated processes, thereby improving modeling efficiency and lowering technical barriers. The AutoML framework and system based on structured data came into being in this context, providing a more convenient and efficient solution for processing and analyzing structured data.

At present, through the analysis and research of existing open source AutoML frameworks, it is found that there are deficiencies in the following aspects: 1) Generality and adaptability. Some AutoML frameworks may have better adaptability to specific types of data or problem areas, but are not flexible enough for other types of data or problems. They cannot cover all possible machine learning scenarios and needs, limiting their generality in diverse applications; 2) Model interpretability. In fields that require understanding of the model decision process (such as finance, medical care, etc.), lack of interpretability may become an obstacle to the adoption of AutoML; 3) Limitations of hyperparameter optimization. Although existing AutoML frameworks provide hyperparameter optimization functions, the search space may still be limited and the global optimal solution cannot be found; 4) Limitations of ensemble learning. Some AutoML systems may not fully utilize ensemble learning methods.

In view of this, how to provide an automatic machine learning method based on structured data that is versatile, flexible, and efficient is a technical problem that needs to be urgently solved by technical personnel in this field.

Summary of the invention

In order to solve the above technical problems, the purpose of the present invention is to provide an automatic machine learning method and system based on structured data, which greatly reduces the workload of manual model selection, while improving the versatility and flexibility of the model in various tasks, and can handle multiple tasks at one time, thereby saving a lot of time and energy and improving efficiency.

The first object of the present invention is to provide an automatic machine learning method based on structured data;

The technical solution provided by the present invention is as follows:

An automatic machine learning method based on structured data comprises the following steps:

Obtaining a structured data set, and labeling a target variable for the structured data;

Performing type identification on the structured data set to obtain a target task type and a target feature type;

Construct corresponding feature engineering according to the target feature type to obtain a target feature combination;

Constructing a base model according to the target task type, and selecting a model fusion strategy corresponding to the base model;

Inputting the target feature combination into the base model for training to obtain a target base model;

The target base model is fused according to the model fusion strategy to obtain a fusion model, and the target base model and the fusion model are automatically deployed.

Preferably, the type identification of the structured data set to obtain the target task type and the target feature type specifically includes:

After removing duplicates from the target variable in the structured data set, the number of categories is counted:

If the number of categories is equal to 2, it is identified as a binary classification task type;

If the data type identified by the target variable is a string and the number of categories counted after deduplication is greater than 2, it is identified as a multi-classification task type;

If the data type of the target variable is an integer or a floating point number, it is identified as a regression task.

Preferably, the type identification of the structured data set to obtain the target task type and the target feature type specifically includes:

Eliminating the target variable from the structured data set to obtain a target feature set;

The target feature set is traversed to obtain the target feature type.

Preferably, before acquiring the target feature combination according to the target feature type, the method further includes:

The target feature type is subjected to data preprocessing.

Preferably, constructing corresponding feature engineering according to the target feature type to obtain a target feature combination specifically includes:

Performing feature transformation according to the target feature type to obtain the target feature;

Performing feature selection on the target feature to obtain a weight score of the target feature;

Traversing the number of each target feature, and performing feature recursive elimination and 5-fold cross validation to obtain an average cross validation score;

The target feature combination is obtained according to the average cross-validation score.

Preferably, the building of a base model according to the target task type specifically includes:

Select a corresponding algorithm as the base model according to the target task type.

Preferably, the step of inputting the target feature combination into the base model for training to obtain a target base model specifically includes:

The target feature combination is input into the list of the base models for parallel training to obtain a target base model.

The second object of the present invention is to provide an automatic machine learning system based on structured data;

The technical solution provided by the present invention is as follows:

An automatic machine learning system based on structured data, comprising: a first acquisition module, a recognition module, a second acquisition module, a construction module, a training module and a deployment module;

The first acquisition module is used to acquire a structured data set and set a target variable for the structured data;

The identification module is used to perform type identification on the structured data set to obtain a target task type and a target feature type;

The second acquisition module is used to construct a corresponding feature engineering according to the target feature type to obtain a target feature combination;

The construction module is used to construct a base model according to the target task type and select a model fusion strategy corresponding to the base model;

The training module is used to input the target feature combination into the base model for training to obtain a target base model;

The deployment module is used to fuse the target base model according to the model fusion strategy to obtain a fusion model, and automatically deploy the target base model and the fusion model.

The third object of the present invention is to provide an electronic device;

The technical solution provided by the present invention is as follows:

An electronic device, comprising:

at least one processor; and

A memory communicatively connected to the at least one processor, the memory storing a computer program executable by the at least one processor, the computer program being executed by the at least one processor so as to enable the at least one processor to perform the method steps described in the automatic machine learning method based on structured data.

A fourth object of the present invention is to provide a computer readable storage medium;

The technical solution provided by the present invention is as follows:

A computer-readable storage medium is used to store a computer program, wherein the computer program is used to cause a computer to execute the method steps described in the automatic machine learning method based on structured data.

The present invention provides an automatic machine learning method based on structured data, comprising the steps of: obtaining a structured data set and labeling the target variables for the structured data; performing type identification on the structured data set to obtain a target task type and a target feature type; constructing a corresponding feature engineering according to the target feature type to obtain a target feature combination; constructing a base model according to the target task type, and selecting a model fusion strategy corresponding to the base model; inputting the target feature combination into the base model for training to obtain a target base model; fusing the target base model according to the model fusion strategy to obtain a fusion model, and automatically deploying the target base model and the fusion model; the method can greatly reduce the workload of manually selecting a model, while improving the versatility and flexibility of the model in various tasks, and can handle multiple tasks at one time, thereby saving a lot of time and energy and improving efficiency.

The present invention also provides an automatic machine learning system based on structured data. Since the system and the automatic machine learning method based on structured data solve the same technical problem and belong to the same technical concept, they should have the same beneficial effects and will not be described in detail here.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

FIG1 is a flow chart of an automatic machine learning method based on structured data in an embodiment of the present invention;

FIG2 is a schematic diagram of the structure of an automatic machine learning system based on structured data in an embodiment of the present invention;

FIG. 3 is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention.

DETAILED DESCRIPTION

In order to enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application are clearly and completely described below. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

It should be noted that when an element is referred to as being "fixed on" or "set on" another element, it can be directly on the other element or indirectly set on the other element; when an element is referred to as being "connected to" another element, it can be directly connected to the other element or indirectly connected to the other element.

It should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, "multiple" and "several" mean two or more, unless otherwise clearly and specifically defined.

It should be noted that the structures, proportions, sizes, etc. illustrated in the drawings of this specification are only used to match the contents disclosed in the specification for people familiar with this technology to understand and read, and are not used to limit the conditions under which this application can be implemented. Therefore, they have no substantive technical significance. Any structural modification, change in proportional relationship or adjustment of size should still fall within the scope of the technical content disclosed in this application without affecting the effects and purposes that can be achieved by this application.

As shown in FIG1 , an embodiment of the present invention provides an automatic machine learning method based on structured data, comprising the following steps:

S1. Obtain a structured data set and annotate the target variable for the structured data;

In step S1, a structured data set is imported through the data import module. There are two ways to import, one is to directly upload through a local file, and the other is to directly read the structured data set of the table in the database. In the process of importing the structured data set, the columns of the target variables need to be marked. If there is a user primary key, the column of the user primary key also needs to be specified. If not, it is not necessary and the system will default to empty.

S2. performing type identification on the structured data set to obtain a target task type and a target feature type;

In step S2, the imported structured data set is input into the automatic identification module to automatically identify the feature type and task type; according to the data analysis of the target variable, the target task type is automatically identified, wherein the target task type includes binary classification, multi-classification and regression; then the target variable and the user primary key are removed, and the remaining is the feature variable, and then the feature variables are traversed in turn, and according to the data distribution, the target feature type of the feature variable is automatically identified, wherein the target feature type includes discrete type, integer type, floating point type, text type, and time type; through this method, the system can automatically analyze the distribution characteristics of the data, so as to determine the type of the data (such as continuous type, discrete type, categorical type, text type, and time type); at the same time, the system can also automatically infer the appropriate task type, such as binary classification task, multi-classification task, and regression task, according to the data distribution of the target variable and the characteristics of the relevant algorithm; this automatic identification process greatly simplifies the data processing process, reduces the interference of human factors, and improves the efficiency of data processing; in addition, since the automatic identification method is based on the statistical characteristics of the data, its identification result is more objective and accurate, which helps to improve the effect of subsequent data analysis and modeling.

S3. Construct corresponding feature engineering according to the target feature type to obtain the target feature combination;

In step S3, the structured data set is processed by feature engineering, different feature transformations are performed according to the target feature type, and then the target feature type after feature transformation is subjected to feature selection, and automatic parameter adjustment and cross-validation are performed to obtain the weight score of the feature, and the weight score is sorted, the number of each feature is traversed, and recursive feature elimination and cross-validation are performed to obtain the average cross-validation score, thereby obtaining the best feature combination; the main idea of recursive feature elimination is to repeatedly build a model, then select the best feature, put the selected feature aside, and then repeat this process on the remaining features until all features are traversed; the feature engineering used in this embodiment is a process of converting raw data into features.

S4. construct a base model according to the target task type, and select a model fusion strategy corresponding to the base model;

In step S4, based on the automatically identified target task type, algorithms corresponding to the target task type are selected to construct base models respectively, and model fusion strategies corresponding to the base models are selected; the base model in this embodiment refers to the selected algorithm list; the model fusion strategies include the Stacking hierarchical integration method and the Voting integration strategy.

S5. Inputting the target feature combination into the base model for training to obtain a target base model;

In step S5, the base models are trained in parallel using the best feature combination and target variable, a set of hyperparameter ranges is set for each base model, and then the optimal parameter combination is searched using GridSearch. Cross-validation (5-fold cross-validation) is used to evaluate the model performance for each base model, and the base model with the best model performance is selected as the target base model.

S6. According to the model fusion strategy, the target base model is fused to obtain a fusion model, and the target base model and the fusion model are automatically deployed.

In step S6, the target base model list is fused or integrated using the Stacking hierarchical integration method and the Voting integration strategy, wherein the second-layer model (meta-model) uses the prediction results of the first-layer model (i.e., the target base model) as input features; the implementation steps are as follows: 1. Train the first-layer models and use them to predict the training data; 2. Use the prediction results of the first-layer models as features to train the second-layer models (meta-models); 3. Use the test data to evaluate the performance indicators of the target base model to obtain a fused model; Voting is a simple integration strategy in which multiple models predict the same data set and select the final prediction result by voting (such as hard voting or soft voting); the implementation steps are as follows: 1. Vote using the selected target base model , and evaluate the performance indicators of the target base model to obtain the fusion model; the target base model and the fusion model are automatically deployed, and an inference service interface is provided for users, so that users can call different models according to the model name during the calling process, and flexible configuration options and extension interfaces are provided to realize the automatic deployment and efficient inference service of the trained model, providing users with a convenient and reliable model inference solution; at the same time, an interactive interface between the user and the system is provided, allowing the user to set the parameters and configuration of the system, allowing the user to select the appropriate algorithm and model structure through the user interface, and providing a visual interface for the user, and being able to view the effect of the model through the page, including the model task type, and the corresponding various evaluation indicators, feature importance, ROC curve, P-R curve.

Compared with the existing technology, this method has the following advantages: 1) versatility and flexibility. It can automatically select suitable algorithms and models, and train and optimize them according to the characteristics of the data. This greatly reduces the workload of manual model selection and parameter adjustment, while improving the versatility and flexibility of the model in various tasks; 2) Lowering the threshold and providing a friendly interface for non-machine learning experts, so that they can easily use structured data for model training and prediction without having to deeply understand the principles and details of machine learning; 3) Improving efficiency. For data scientists and machine learning engineers, there is no need to design and adjust models for different types of tasks separately. The design-based automated machine learning method can handle multiple tasks at one time, saving a lot of time and energy.

Preferably, the type identification of the structured data set to obtain the target task type and the target feature type specifically includes:

After removing duplicates from the target variable, the number of categories is counted:

If the number of categories is equal to 2, it is identified as a binary classification task type;

If the data type identified by the target variable is a string and the number of categories counted after deduplication is greater than 2, it is identified as a multi-classification task type.

If the data type of the target variable is an integer or a floating point number, it is identified as a regression task.

In actual application, the target task type is automatically identified based on the data analysis of the target variable, and the target task types include binary classification, multi-classification and regression. The identification standard of the target task type is as follows: after deduplication of the target variable, the number of categories is counted. If the number of categories is equal to 2, it is identified as a binary classification task. If the data type identified by the target variable is a string and the number of categories counted after deduplication is greater than 2, it is identified as a multi-classification task. If the data type identified by the target variable is an integer or a floating-point number, it is identified as a regression task.

Preferably, the type identification of the structured data set to obtain the target task type and the target feature type specifically includes:

Eliminating the target variable from the structured data set to obtain a target feature set;

The target feature set is traversed to obtain the target feature type.

In actual application, the target variable and the user primary key are removed, and the remaining ones are the feature variables. Then the feature variables are traversed in turn, and the target feature type of the feature variable is automatically identified according to the data distribution. The target feature types include discrete, integer, floating-point, text and time types. The recognition standard for discrete type: the data type is object, bool or category, and the field length is less than or equal to 25; the recognition standard for integer type: the data type contains "int"; the recognition standard for floating-point type: the data type includes "float"; the recognition standard for text type: the data type is object, bool or category, and the field length is greater than 25; the recognition standard for time type: the data type is object or category, and can be processed by the to_datetime function in the pandas library.

Preferably, before acquiring the target feature combination according to the target feature type, the method further includes:

The target feature type is subjected to data preprocessing.

In the actual application process, the feature data source in the target feature type is preprocessed. First, the null value rate of each feature is counted, and the columns with null value rates exceeding 80% are filtered out and removed. Secondly, missing values are processed. For discrete features, "-99999" is used to fill missing values; for integer features, 0 is used to fill missing values; for floating-point features, the average value of the feature column is used to fill missing values; for text features, empty strings are used to fill missing values; for time features, the previous row of the feature column is used to fill missing values.

Preferably, constructing corresponding feature engineering according to the target feature type to obtain a target feature combination specifically includes:

Performing feature transformation according to the target feature type to obtain the target feature;

Performing feature selection on the target feature to obtain a weight score of the target feature;

Traversing the number of each target feature, and performing feature recursive elimination and 5-fold cross validation to obtain an average cross validation score;

The target feature combination is obtained according to the average cross-validation score.

In the actual application process, the target feature type after data preprocessing is used to construct feature engineering. First, for discrete feature columns, One-Hot encoding is performed on them, using OneHotEncoder in the sklean.preprocessing library; for floating-point features, they are normalized, using StandardScaler in the sklean.preprocessing library; for integer features, they are binned, using KBinsDiscretizer in the sklean.preprocessing library, where the number of bins is calculated using the Sturgess formula, which is num_bins=int(1+np.log2(n), where int is rounded, and np.log2 is NumPy (a Python library). n) is a function used to calculate the logarithm with base 2; for time-type features, the year, month, day, hour, week, day of the week, whether it is a working day, whether it is the beginning of the month, and whether it is the end of the month are extracted as new feature columns; for text-type features, TfidfVectorizer in sklearn.feature_extraction.text is used to implement it; then feature selection is performed based on the feature list after feature transformation, and feature selection is based on random forest and automatic parameter adjustment and 5-fold cross-validation based on the transformed features to obtain the weight score of the feature, and sort it, traverse each feature number, and perform RFE and 5-fold cross-validation, store the average cross-validation score, and obtain the best feature combination, that is, the target feature combination; the importance of weights and the feature selection process are visualized.

Preferably, the building of a base model according to the target task type specifically includes:

Select a corresponding algorithm as the base model according to the target task type.

In actual application, based on the automatically identified task type, the corresponding algorithm is selected as the base model. For example, for binary classification tasks, five classifiers, RF (random forest), LightGBM (lightweight gradient boosting machine), AdaBoost (adaptive boosting), XGBoost (extreme gradient boosting tree) and LR (logistic regression) are selected as base models; for multi-classification tasks, five classifiers, RF (random forest), LightGBM (lightweight gradient boosting machine), AdaBoost (adaptive boosting), XGBoost (extreme gradient boosting tree) and LR (logistic regression) are selected as base models; for regression tasks, three regression algorithms, RF (random forest), LightGBM (lightweight gradient boosting machine) and XGBoost (extreme gradient boosting tree) are selected as base models.

Preferably, the step of inputting the target feature combination into the base model for training to obtain a target base model specifically includes:

The target feature combination is input into the list of the base models for parallel training to obtain a target base model.

In the actual application process, the base model is trained based on the selected feature list and each base model is cross-validated by 5 folds. The hyperparameter optimization unit is combined to automatically tune the hyperparameters of each base model. A set of hyperparameter ranges is set for the base model in each model task, which can automatically adjust the parameters and structure of the model according to the characteristics of the task, thereby optimizing the performance of the model and ensuring that the effect of manual parameter adjustment can be achieved or close to that of manual parameter adjustment on various tasks. The optimal parameter combination is searched based on the GridSearch grid search method, and the target feature combination is input into the base model for training. Each base model is evaluated and the optimal base model is output. The performance indicators of the binary classification task include accuracy, precision, recall, F1-Score, area under curve (AUC), receiver operating characteristic curve (ROC) curve, P-R curve and confusion matrix; the evaluation indicators of the multi-classification task include accuracy, precision, recall, F1-Score, area under curve (AUC), receiver operating characteristic curve (ROC) curve, P-R curve and confusion matrix. The performance indicators of the regression task include mean absolute error (MAE), mean square error (MSE), mean absolute percentage error (MAPE), and regression score function (R2 Score determination coefficient).

As shown in FIG2 , the present invention also provides an automatic machine learning system based on structured data, comprising: a first acquisition module, a recognition module, a second acquisition module, a construction module, a training module, and a deployment module;

The first acquisition module is used to acquire a structured data set and set a target variable for the structured data;

The identification module is used to perform type identification on the structured data set to obtain a target task type and a target feature type;

The second acquisition module is used to construct a corresponding feature engineering according to the target feature type to obtain a target feature combination;

The construction module is used to construct a base model according to the target task type and select a model fusion strategy corresponding to the base model;

The training module is used to input the target feature combination into the base model for training to obtain a target base model;

The deployment module is used to fuse the target base model according to the model fusion strategy to obtain a fusion model, and automatically deploy the target base model and the fusion model.

In the actual application process, a first acquisition module, an identification module, a second acquisition module, a construction module, a training module and a deployment module are set; the identification module is connected to the first acquisition module, the second acquisition module and the construction module respectively; the training module is connected to the second acquisition module, the construction module and the deployment module respectively; the first acquisition module acquires a structured data set, sets target variables for the structured data, and then sends the structured data set to the identification module; the identification module performs type identification on the structured data set to acquire the target task type and the target feature type, and then sends the target task type to the construction module, and sends the target feature type to the second acquisition module; the second acquisition module acquires the target task type according to the target feature type, and then sends the target feature type to the second acquisition module. After obtaining the target feature combination, the target feature combination is sent to the training module; after the construction module constructs the base model according to the target task type, the base model is sent to the training module; after the training module inputs the target feature combination into the base model for training to obtain the target base model, the target base model is sent to the deployment module; the deployment module fuses the target base model according to the model fusion strategy to obtain the fusion model, and automatically deploys the target base model and the fusion model; this system can greatly reduce the workload of manual model selection, while improving the versatility and flexibility of the model in various tasks, and can handle multiple tasks at one time, thereby saving a lot of time and energy and improving efficiency.

Furthermore, an embodiment of the present application also discloses an electronic device. FIG3 is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content in the diagram cannot be regarded as any limitation on the scope of use of the present application.

FIG3 is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input/output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the automatic machine learning method based on structured data disclosed in any of the aforementioned embodiments. In addition, the electronic device 20 in this embodiment may specifically be an electronic computer.

In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.

In addition, the memory 22, as a carrier for storing resources, can be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon may include an operating system 221, a computer program 222 and data 223, etc. The storage method can be temporary storage or permanent storage.

Among them, the operating system 221 is used to manage and control the hardware devices and computer programs 222 on the electronic device 20 to realize the operation and processing of the data 223 in the memory 22 by the processor 21, which can be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program that can be used to complete the automatic machine learning method based on structured data performed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 222 can further include a computer program that can be used to complete other specific tasks. In addition to data transmitted from an external device received by the automatic machine learning device based on structured data, the data 223 can also include data collected by its own input and output interface 25, etc.

The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

Furthermore, the present application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the automatic machine learning method based on structured data disclosed above is implemented. The specific steps of the method can be referred to the corresponding contents disclosed in the above embodiments, and will not be repeated here.

In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

If a flow chart is used in the present application, the flow chart is used to illustrate the operations performed by the system according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, each step can be processed in reverse order or simultaneously. At the same time, other operations can also be added to these processes, or a certain step or several steps of operations can be removed from these processes.

The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims (10)

1. An automatic machine learning method based on structured data, comprising the steps of:

Obtaining a structured data set, and labeling target variables for the structured data;

performing type recognition on the structured data set to acquire a target task type and a target feature type;

constructing corresponding feature engineering according to the target feature type to obtain a target feature combination;

constructing a base model according to the target task type, and selecting a model fusion strategy corresponding to the base model;

Inputting the target feature combination into the base model for training to obtain a target base model;

And fusing the target base model according to the model fusion strategy to obtain a fusion model, and automatically deploying the target base model and the fusion model.

2. The structured data based automatic machine learning method of claim 1 wherein said type identifying said structured dataset to obtain a target task type and a target feature type comprises:

After the target variable in the structured dataset is de-duplicated, counting the category number:

If the category number is equal to 2, identifying the task as a category task type;

if the data type identified by the target variable is a character string and the counted class number after duplication removal is more than 2, identifying the data type as a multi-classification task type;

And if the data type identified by the target variable is an integer or a floating point number, identifying the data type as a regression task.

3. The structured data based automatic machine learning method of claim 1 wherein said type identifying said structured dataset to obtain a target task type and a target feature type comprises:

Removing the target variable from the structured dataset to obtain a target feature set;

traversing the target feature set to obtain the target feature type.

4. The structured data based automatic machine learning method of claim 1 wherein prior to said obtaining a target feature combination from said target feature type, further comprising:

And carrying out data preprocessing on the target feature type.

5. The structured data based automatic machine learning method of claim 1, wherein said constructing a corresponding feature project according to the target feature type to obtain a target feature combination specifically comprises:

Performing feature transformation according to the target feature type to obtain target features;

Performing feature selection on the target feature to obtain a weight score of the target feature;

traversing the number of each target feature, and performing feature recursion elimination and 5-fold cross validation to obtain an average cross validation score;

And obtaining the target feature combination according to the uniform cross validation score.

6. The structured data based automatic machine learning method of claim 1 wherein said constructing a base model from said target task type comprises:

And selecting a corresponding algorithm as a base model according to the target task type.

7. The structured data based automatic machine learning method of claim 1, wherein said inputting said target feature combinations into said base model for training to obtain a target base model, in particular comprising:

and inputting the target feature combination into the list of the base models for parallel training so as to acquire target base models.

8. An automated machine learning system based on structured data, comprising: the system comprises a first acquisition module, an identification module, a second acquisition module, a construction module, a training module and a deployment module;

The first acquisition module is used for acquiring a structured data set and setting a target variable for the structured data;

the identification module is used for carrying out type identification on the structured data set so as to acquire a target task type and a target feature type;

the second obtaining module is used for constructing corresponding feature engineering according to the target feature type so as to obtain a target feature combination;

The construction module is used for constructing a base model according to the target task type and selecting a model fusion strategy corresponding to the base model;

The training module is used for inputting the target feature combination into the base model for training so as to obtain a target base model;

The deployment module is used for fusing the target base model according to the model fusion strategy to obtain a fusion model, and automatically deploying the target base model and the fusion model.

9. An electronic device, comprising:

at least one processor; and

A memory communicatively coupled to the at least one processor, the memory storing a computer program executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 7.

10. A computer readable storage medium, characterized in that the storage medium is for storing a computer program for causing a computer to execute the method of any one of claims 1 to 7.