CN116028881B - Method and system for predicting recurrence of restrictive behavior based on multidimensional database - Google Patents

Abstract

The invention discloses a prediction method, system, computer equipment, and storage medium for reoccurrence restriction behavior based on a multidimensional database. The realization of the method includes: constructing a multidimensional heterogeneous database; generating a sample data set, the sample data set includes a training data set and Test data set; according to the restricted behavior data of people with restricted activities within the preset period, divide positive and negative cases as positive and negative case labels; build the original short-term and short-term recurrence restricted behavior risk prediction model; through association query in The feature vector group of preset dimensions is selected in the training data set as input data, which is input to the original short-term re-occurrence restriction behavior risk prediction model for prediction processing, so as to output the prediction probability of re-appearance restriction behavior; calculate the loss function value, according to The loss function value iteratively trains the original short-term and short-term re-emergence behavior risk prediction model until the preset conditions are met, and generates the actual short-term and medium-term re-emergence restriction behavior risk prediction model. It can effectively improve the accuracy of forecasting and improve the overall efficiency of personnel management in places where activities are restricted.

Description

Method and system for predicting recurrence of restrictive behavior based on multidimensional database

technical field

The present invention relates to the technical field of deep neural networks, and in particular to a prediction method, system, computer equipment and storage medium based on a multidimensional database-based recurrence limiting behavior.

Background technique

The management and transformation of personnel with restricted activities is a very important issue, whether it is in the academic circle or at the national management level. How to discover and regulate the risk of restricted behaviors again under the premise of humanitarianism is particularly important, and in the restricted activities The social network in the venue is an important unfavorable factor that hinders the transformation of restricted activity personnel in restricted activity venues, and it is also an important risk variable for the cross-interference of restricted activity personnel in restricted activity venues. However, the current traditional method is to rely on supervisors to detect abnormalities of restricted behavior personnel through video, conversation, and manual work before taking precautions. Not only is it inefficient and labor-intensive, it cannot make full use of the high-risk social networks in restricted activity venues, but also has low accuracy and unscientific problems due to the manual method of predicting the probability of reoccurrence of restricted behaviors.

Contents of the invention

Based on this, it is necessary to provide a prediction method, system, computer equipment and storage medium based on a multi-dimensional database based on the above-mentioned technical problems, so as to solve at least one problem existing in the above-mentioned prior art.

In the first aspect, a method for predicting the recurrence of limiting behavior based on a multidimensional database is provided, including:

Constructing a multi-dimensional heterogeneous database, the multi-dimensional heterogeneous database includes the data of the previous legal documents, the relevant data of the previous entry into the restricted activity site, the monitoring data in the previous restricted activity site, and the legal document data of the current restricted behavior;

Perform data mining and data cleaning on the data of the previous legal documents, the relevant data of the previous entry into the restricted activity place, the monitoring data in the previous restricted activity place, and the legal document data of the current restricted behavior to generate a sample data set. The sample data set includes a training data set and a testing data set;

According to the restricted behavior data of the restricted activities personnel reappearing within the preset period, positive and negative cases are divided as positive and negative case labels;

Construct the original short-to-medium-term reappearance restriction behavior risk prediction model;

Selecting a eigenvector group of preset dimensions in the training data set through an associated query, as input data, inputting it into the original medium and short-term re-occurrence restriction behavior risk prediction model for prediction processing, so as to output the re-appearance restriction behavior prediction probability;

Calculate the loss function value according to the prediction probability of re-occurrence restriction behavior and the positive and negative example labels, and iteratively train the original medium-term and short-term re-appearance restriction behavior risk prediction model according to the loss function value until the preset condition is met. , generating an actual short-to-medium-term risk prediction model for re-emergence of restricted behaviors, so as to predict the risk of restricted behaviors re-emerged within a preset time range by persons with restricted activities through the actual short-to-medium-term risk prediction model for re-emergence of restricted behaviors.

In one embodiment, after the multidimensional heterogeneous database is constructed, it includes:

Obtain the parameters of intimacy between persons with different restricted activities by performing video image analysis on the monitoring data in the previous restricted activity place;

According to the intimacy parameter, calculate the intimacy with other persons who jointly participate in the restrictive behavior as the data to be predicted, and classify it into the sample data set.

In one embodiment, according to the restricted activity data of the restricted activity personnel again within the preset period, the corresponding restricted activity personnel in the sample data set are divided into positive and negative cases, including:

When the restrictive behavior event occurs again within the preset time limit for the restricted activity personnel, classify the corresponding restricted activity personnel in the sample data set as a negative example;

When the person with restricted activities does not reappear the restricted behavior event within the preset time limit, the corresponding person with restricted activities in the sample data set is classified as a positive case.

In one embodiment, the calculation of the loss function value according to the predicted probability of reoccurrence of the restricted behavior and the positive and negative labels includes:

By using a preset loss function and a backpropagation algorithm, the value of the loss function is calculated for the predicted probability of the recurrence restriction behavior and the labels of positive and negative examples.

In one embodiment, after generating the actual short-to-medium-term reoccurrence restriction behavior risk prediction model, it includes:

Obtain the legal document data of the current restricted behavior of the person to be predicted to restrict activities, the relevant data of entering the restricted activity place due to the restricted behavior, and the life data in the restricted activity place;

Perform data mining and data cleaning on the legal document data of the current restricted behavior, the relevant data of entering the restricted activity place due to the restricted behavior, and the life data in the restricted activity place, so as to generate the characteristic data to be predicted;

The to-be-predicted feature data is input into the actual short-to-medium-term recurrence risk prediction model to output the prediction probability of the recurrence restriction behavior of the person to be predicted to restrict activities.

In one embodiment, the actual short-to-medium-term reappearance restrictive behavior risk prediction model includes an input layer, a plurality of hidden layers, and an output layer, and the characteristic data to be predicted is input to the actual short-term re-emergence restrictive behavior The risk prediction model is used to output the predicted probability of the restricted behavior of the person to be predicted to be restricted, including:

receiving the feature data to be predicted through the input layer;

Inputting the feature data to be predicted received by the input layer into multiple hidden layers in sequence for processing;

The output layer outputs the prediction probability of the reappearance restriction behavior of the person to be predicted and obtained after the last layer of hidden layer processing.

In an embodiment, after outputting the predicted probability of the re-emergence of restricted behaviors of the person to be predicted based on the actual short-term and short-term re-emergence restricted behavior risk prediction model, it includes:

When the predicted probability of the reoccurrence of the restricted behavior of the person to be predicted with restricted activities is greater than the preset threshold, the risk of the reappeared restricted behavior of the person to be predicted with restricted activities is visualized, and the key observation is sent to the restricted activity place management platform list.

In the second aspect, a multidimensional database-based prediction system for reoccurrence of limiting behavior is provided, including:

The database construction unit is used to construct a multi-dimensional heterogeneous database, which includes the data of the previous legal documents, the relevant data of the previous entry into the restricted activity site, the monitoring data in the previous restricted activity site, and the current restricted behavior. legal document data;

a data set generation unit, configured to perform data mining and data cleaning on the previous legal document data, the relevant data of the previous entry into the restricted activity place, the monitoring data in the previous restricted activity place, and the legal document data of the current restricted behavior, To generate a sample data set, the sample data set includes a training data set and a testing data set;

The label generation unit is used to divide the positive and negative examples according to the restricted behavior data of the restricted activity personnel within the preset period, so as to serve as positive and negative example labels;

A model construction unit, used to construct the original short-to-medium-term reappearance restriction behavior risk prediction model;

The training unit is used to select a feature vector group of preset dimensions in the training data set through an associated query, as input data, and input it into the original short-term recurrence restriction behavior risk prediction model for prediction processing, and output the recurrence Constraining behavioral prediction probabilities;

The prediction unit is used to calculate the loss function value according to the predicted probability of re-occurrence restriction behavior and the positive and negative example labels, and iteratively train the original short-term re-emergence restriction behavior risk prediction model according to the loss function value until When the preset conditions are met, an actual medium-to-short-term re-emergence risk prediction model of restriction behavior is generated, so as to predict the risk of re-emergence of restriction behavior within the preset time range of restricted activity personnel through the actual medium-to-short-term re-emergence risk prediction model.

In a third aspect, a computer device is provided, including a memory, a processor, and computer-readable instructions stored in the memory and operable on the processor, when the processor executes the computer-readable instructions The steps of implementing the method for predicting the reoccurrence limit behavior based on the multidimensional database as described above.

In a fourth aspect, a readable storage medium is provided, the readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the recurrence restriction behavior based on the multidimensional database as described above is implemented The steps of the forecasting method.

The above-mentioned prediction method, system, computer equipment and storage medium based on the multi-dimensional database-based re-occurrence restriction behavior, the realization of the method includes: constructing a multi-dimensional heterogeneous database, the multi-dimensional heterogeneous database includes the previous legal document data, the previous entry Relevant data of restricted activity places, monitoring data in the previous restricted activity places, and legal document data of current restricted activities; Data mining and data cleaning of monitoring data and legal document data currently restricting behaviors to generate sample data sets, the sample data sets include training data sets and test data sets; Restrict behavioral data, and divide positive and negative cases as positive and negative case labels; construct the original short-term and short-term reappearance restriction behavior risk prediction model; select a feature vector group with preset dimensions in the training data set through correlation query as input The data is input into the original short-to-medium-term recurrence risk prediction model for prediction processing, so as to output the predicted probability of the restricted behavior again; according to the predicted probability of the restricted behavior again and the positive and negative example labels, the loss function value is calculated , according to the loss function value, iteratively train the original short-term reappearance restriction behavior risk prediction model until the preset conditions are met, and generate an actual short-term re-appearance restriction behavior risk prediction model to pass the actual short-term reappearance restriction behavior risk prediction model. The risk prediction model for the recurrence of restricted behaviors predicts the risk of restricted behaviors reappearing within the preset time range of restricted activity personnel. In this application, by analyzing public legal documents, personal family information of persons with restricted activities, social life data in places with restricted activities, etc., combined with deep neural network, and by dividing positive and negative cases of persons within a specified period, a middle-level The short-term prediction model for the reoccurrence of restricted behaviors of restricted behavior personnel predicts the probability of restricted behaviors reoccurring within the preset period, and can be combined with a visualization platform to provide key observation lists for restricted activity site managers to improve restrictions Overall efficiency of event venue personnel management. Effectively improve the accuracy and efficiency of forecasting.

Description of drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings that need to be used in the description of the embodiments of the present invention. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention , for those skilled in the art, other drawings can also be obtained according to these drawings without paying creative labor.

FIG. 1 is a schematic diagram of an application environment of a prediction method based on a multidimensional database-based recurrence restriction behavior in an embodiment of the present invention;

FIG. 2 is a schematic flow chart of a method for predicting reoccurrence restriction behavior based on a multidimensional database in an embodiment of the present invention;

Fig. 3 is a schematic diagram of the forecasting process of the short-term and medium-term reappearance restriction behavior risk forecasting model in an embodiment of the present invention;

Fig. 4 is a schematic structural diagram of a risk prediction model for short-term and medium-term re-emergence of restricted behaviors in an embodiment of the present invention;

FIG. 5 is a structural schematic diagram of a method for predicting reoccurrence restriction behavior based on a multidimensional database in an embodiment of the present invention;

FIG. 6 is a schematic diagram of computer equipment in an embodiment of the present invention.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some of the embodiments of the present invention, but not all of them. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

The method for predicting the re-emergence of restrictive behavior based on the multidimensional database provided in this embodiment can be applied in the application environment as shown in Figure 1, wherein the multi-dimensional heterogeneous database is used to provide training sample data for the short-term re-emergence of restrictive behavior risk prediction model And test sample data, the short-to-medium-term reappearance of restricted behavior risk prediction model after the training and testing of the training sample data and test sample data, the legal document data of the current restricted behavior of the restricted activities, the restricted activities of the restricted activities of the personnel due to the restricted behavior Predict the risk of reoccurrence of restricted behaviors based on the information of entering the restricted activity sites and the life information in the restricted activity site, and output the prediction results to the visualization platform for the reappearance of restricted behavior risks through the network, etc., to improve the management personnel supervision of restricted activity sites accuracy and efficiency.

Among them, re-emerging restricted conduct risk visualization platforms include, but are not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The short-to-medium-term reappearance restriction behavior risk prediction model can be set on the server side, and can also be set on the local monitoring and management platform. The server side can be realized by an independent server or a server cluster composed of multiple servers.

In one embodiment, as shown in FIG. 2 , a method for predicting the reoccurrence restriction behavior based on a multidimensional database is provided, including the following steps:

In step S110, a multi-dimensional heterogeneous database is constructed, the multi-dimensional heterogeneous database includes the data of the previous legal documents, the relevant data of the previous entry into the restricted activity place, the monitoring data in the previous restricted activity place, and the current law of restrictive behavior. document data;

Through the public document database or cooperation with the restricted activity site management platform, legal document management platform, etc., to obtain the original criminal judgment document data, relevant information and data collected when the restricted activity personnel in the restricted activity place enter the restricted activity place, Surveillance video data during the period of restricted activities, as well as document data on the recurrence of restricted behaviors, etc., together constitute a multi-source heterogeneous database of restricted activities, which is used to provide information for the construction of subsequent risk prediction models for short-term and medium-term re-emergence of restricted behaviors. sample.

Among them, the person whose activities are restricted refers to the person whose activities are restricted due to the implementation of restricted behavior, which can be understood as an act that violates the law. The restricted activity place can be a place where the restricted activity personnel who will have restricted activities are managed in a centralized manner and the scope of activities of the personnel is restricted.

Among them, the data of the previous legal document, the relevant data of the previous entry into the restricted activity place, and the monitoring data in the previous restricted activity place can specifically be the legal document data generated when the user’s last restricted behavior occurred, as well as the And enter the information data of restricted activity places and the monitoring data in restricted activity places.

In step S120, data mining and data cleaning are performed on the previous legal document data, the relevant data of the previous entry into the restricted activity place, the monitoring data in the previous restricted activity place, and the legal document data of the current restricted behavior to generate A sample data set, the sample data set includes a training data set and a testing data set;

In the embodiment of this application, after the multi-dimensional heterogeneous database is constructed, due to the complexity of the data in the multi-dimensional heterogeneous database and the existence of unstructured data, unstructured text data, such as legal documents, are restricted due to The information data that enters the restricted activity place, etc., can be mined through feature engineering or deep learning algorithm to extract the available data to be detected, for example, text-CNN, NLP technology, etc., to extract the following data to be detected, for example, Judging whether multiple people implement restricted behavior data, whether to implement emerging restricted behavior, parental related conditions, education level, occupation, etc.

Further, for video surveillance data, video image analysis technology can also be used to perform video analysis on video surveillance data, such as face recognition, image segmentation, object detection, etc., to obtain the required data dimensions, such as participating in the next Restrict behavioral personnel's time together, conflicts, etc.

In this embodiment of the application, when data mining the previous legal document data, the main extracted content may include the cause of action corresponding to the previous restrictive behavior data document, the place where the case occurred, whether multiple people implemented the restrictive behavior, whether the emerging restrict behavior.

In the embodiment of this application, the relevant data of the previous entry to the restricted activity place refers to the relevant data that the restricted activity personnel in the restricted activity place need to fill in and collect when entering the restricted activity place. During data mining, the data that can be extracted Specifically, the data content may include data dimensions divided into two parts: the first part is: family information; the second part is: personal information.

In the embodiment of this application, when data mining is performed on the monitoring data in the previous restricted activity place, the data that can be mined is mainly divided into two parts. The first part is: structured data commonly used in restricted behavior places, such as restricted The number of dangerous behaviors, the number of conflicts, and the commendation situation during the last restricted behavior place of the restricted activity personnel in the activity place. The second part is: after the restricted behavior occurs again, it is necessary to use the previously stored video source data to analyze the video image to obtain quantitative data, such as the time the party spends with other people who participated in the restricted behavior.

Further, for the remaining video source data, the data to be predicted can be obtained in the following ways:

Obtain the parameters of intimacy between persons with different restricted activities by performing video image analysis on the monitoring data in the previous restricted activity place;

According to the intimacy parameter, calculate the intimacy of other persons who jointly participate in the restricted behavior as the data to be predicted, and classify it into the sample data set.

Among them, the storage time of video source files in ordinary places with restricted activities is relatively short, generally three months. Therefore, in order to study the risk of restricted behaviors recurring in a longer period of time, the monitoring data can be the monitoring data stored within a preset time . For example, within 1 year.

Further, by performing video image analysis on the monitoring data in the previous restricted activity place, such as face recognition, object recognition, image segmentation and other technical means, the users appearing in the monitoring data can be identified, and records can be obtained Parameters of intimacy between the user and people with different restricted activities, such as time spent together, conflicts, close communication, etc.

，其中t为当事人与其他共同参与限制行为的人员相处的时间t小时，k为可变超参数，偏置量，用于控制前值的影响程度，初步设定为0.5。从而可计算得到其他共同参与限制行为的人员因密度，并可作为待预测数据，因为限制活动人员在限制活动场所中会有很多交流和社交，并且容易产生一些关系和限制期满，离开限制活动场所后，与其他人员共同再次实施限制行为的情形，因此通过对限制活动场所的监控数据进行视频分析得到限制活动人员之间的亲密关系，以便在通过中短期再次出现限制行为风险预测模型时，可充分考虑高危社交网络。Since the main dependent variable of the degree of intimacy is the next communication time t with other people who participate in the restricted behavior together, and the amount of information during the time spent together does not increase linearly, so the concept of information entropy can be used to set the degree of intimacy as the formula:

, where t is the time t hours that the party spends with other people who participated in the restricted behavior together, and k is a variable hyperparameter, bias, which is used to control the degree of influence of the previous value, and is initially set to 0.5. In this way, the density of other people who jointly participate in restricted behaviors can be calculated, and can be used as data to be predicted, because people with restricted activities will have a lot of communication and social interaction in restricted activity places, and it is easy to have some relationships and leave restricted activities when the restriction expires Therefore, through video analysis of the monitoring data of the restricted activity place, the intimate relationship between the restricted activity personnel can be obtained, so that when the risk prediction model for the restricted behavior reappears in the short and medium term, High-risk social networks can be given full consideration.

In the embodiment of the present application, when data mining is performed on the clerical data of persons who currently have restricted behaviors again, the mined data may include the cause of the case, the place where the case occurred, whether multiple people carried out the restricted behavior, and whether they participated in the restricted behavior with other people. Personnel have the experience of being locked up in a restricted activity place together, whether it is an emerging restricted behavior, etc. And this part of the data can also map the people who shared the restricted behavior data in the video in the previous restricted activity place, and the influence of the intimacy of other people who jointly participated in the restricted behavior generated by the social network in the restricted activity place on the reoccurrence of restricted behavior.

In step S130, according to the restricted behavior data of the restricted activity personnel within the preset time limit, positive and negative cases are divided as positive and negative case labels;

In the embodiment of the present application, the preset time limit may be 1 year, or other time limit, for example, 3 years, 3 months, etc., preferably 1 year.

In this embodiment of the application, the data on the reappearance and restricted behavior of the person with restricted activities within the preset time limit can be obtained by continuously monitoring the status of the person with restricted activities within the preset time limit to obtain the reappearance of the person with restricted activities Limit behavioral data.

In the embodiment of the present application, after obtaining the restricted behavior data of the restricted activity personnel within the preset period, the data can be compared with the pre-stored data in the multi-dimensional heterogeneous database to determine the The data belonging to the person with restricted activities in the multi-dimensional heterogeneous database, that is, to determine whether the person with previously restricted activities has restricted behavior again within the preset time range, and based on this situation, positive and negative Example division.

Further, according to the reoccurrence of the restrictive behavior of the restricted activity personnel within the preset period, the corresponding restricted activity personnel in the sample data set are divided into positive and negative cases, including:

When the restrictive behavior event occurs again within the preset time limit for the restricted activity personnel, classify the corresponding restricted activity personnel in the sample data set as a negative example;

When the person with restricted activities does not reappear the restricted behavior event within the preset time limit, the corresponding person with restricted activities in the sample data set is classified as a positive case.

Specifically, when the previously detained persons with restricted activities reappear within the preset time limit, for example, within one year, the corresponding restricted persons in the sample data set are classified as negative cases, when the restricted When the restricted behavior event does not occur again within the preset period, for example, within 1 year, the corresponding restricted activity personnel in the sample data set are classified as positive cases. By setting the positive and negative examples, that is, the data information that will not appear again in one year and will appear again in the restricted behavior needs to be learned to do the classification task, so as to realize the risk prediction model for the short-term and medium-term reappearance of the training. Supervised learning to improve predictive accuracy.

In step S140, constructing the original short-to-medium-term recurrence risk prediction model of restrictive behavior;

In the embodiment of the present application, the original medium- and short-term reappearance restriction behavior risk prediction model can be constructed by a DNN neural network. The DNN neural network is a multi-layer perceptron. After receiving a signal by simulating the animal nerve, it will send each neuron. After receiving the input, each neuron will activate and generate the input signal according to its own judgment, and then summarize it to realize the identification and classification of the information source. role. The structure is simple, and highly complex functions can be fitted.

Among them, referring to Fig. 3, the original short-to-medium-term recurrence restriction behavior risk prediction model may specifically include: an input layer, multiple hidden layers, and an output layer. The hidden layer may include 4 layers, each of which may be the first 32-dimensional node The hidden layer hidden layer1, the second hidden layer hidden layer2 with 64 nodes, the third hidden layer hidden layer3 with 32 nodes, the fourth hidden layer hidden layer4 with 16 nodes, and the activation function used between each level is the sigmod function.

In step S150, a eigenvector group of preset dimensions is selected in the training data set through an associated query, as input data, and input into the original medium and short-term re-occurrence restriction behavior risk prediction model for prediction processing, to output re-occurrence Constraining behavioral prediction probabilities;

In the embodiment of the present application, the original medium-to-short-term recurrence restriction behavior risk prediction model can be connected to the multi-dimensional heterogeneous database, and a feature vector group of preset dimensions can be selected from the training data set through correlation query as input The data are input into the original short-term recurrence restriction behavior risk prediction model for prediction processing.

Wherein, the feature vector group of preset dimensions may specifically include feature vectors of 40 dimensions. By selecting 40-dimensional feature vectors as the input of model training, since there is a causal relationship between the selected data dimensions and the prediction results, it can provide better prediction results and higher prediction accuracy.

In the embodiment of the present application, a method of associative query may be used to select multi-dimensional feature vectors in multi-source heterogeneous databases. Because in a single data dimension, the relevant information about the last entry into the restricted activity place due to the restricted behavior, the social information of the last restricted activity place and other restricted activity personnel in the restricted activity place need to be obtained by using the associated query. For example, when the restricted activity person A enters the restricted activity place again, the relevant information of the restricted activity person A in the restricted activity place last time can be extracted through associated query, which can specifically include the restricted activity person A Closely communicate relevant information with those people whose activities are restricted, the number of dangerous behaviors and the number of conflicts in the place where activities were restricted last time, etc.

In step S160, the loss function value is calculated according to the predicted probability of re-emergence of restrictive behavior and the labels of positive and negative examples, and the original medium- and short-term re-emergence of restrictive behavior risk prediction model is iteratively trained according to the loss function value until When the preset conditions are met, an actual medium-to-short-term re-emergence risk prediction model of restriction behavior is generated, so as to predict the risk of re-emergence of restriction behavior within the preset time range of restricted activity personnel through the actual medium-to-short-term re-emergence risk prediction model.

In an embodiment of the present application, the calculation of the loss function value based on the predicted probability of reoccurrence of the restricted behavior and the labels of positive and negative examples includes:

By using a preset loss function and a backpropagation algorithm, the value of the loss function is calculated for the predicted probability of the recurrence restriction behavior and the labels of positive and negative examples.

Wherein, the preset loss function may be a cross-entropy function, or a mean square error, KL divergence, and the like. Specifically, it can be set according to the actual situation, which is not limited in this application.

In the embodiment of the present application, the loss function value is calculated for the predicted probability of re-occurrence restriction behavior and the labels of positive and negative examples output by the original short-term re-emergence restriction behavior risk prediction model through the loss function and backpropagation algorithm until The accuracy rate of the loss function value reaches a preset threshold, such as 98%, at this time, an actual short-to-medium-term reappearance restriction behavior risk prediction model can be formed.

Further, after obtaining the actual short-to-medium-term recurrence risk prediction model, including:

After the generation of the actual short-to-medium-term reoccurrence limit behavior risk prediction model, it includes:

Obtain the legal document data of the current restricted behavior of the person to be predicted to restrict activities, the relevant data of entering the restricted activity place due to the restricted behavior, and the life data in the restricted activity place;

Perform data mining and data cleaning on the legal document data of the current restricted behavior, the relevant data of entering the restricted activity place due to the restricted behavior, and the life data in the restricted activity place, so as to generate the characteristic data to be predicted;

The to-be-predicted feature data is input into the actual short-to-medium-term recurrence risk prediction model to output the prediction probability of the recurrence restriction behavior of the person to be predicted to restrict activities.

Specifically, after training the original short-term and short-term reappearance restriction behavior risk prediction model through the sample data in the multi-source heterogeneous database, a short-term and medium-term reappearance restriction behavior risk prediction model that is actually used for prediction can be formed. At this time, the legal document data, restricted behavior data, and life data in the restricted activity place of the person to be predicted to be restricted can be obtained, and data mining and data cleaning are performed to form data that can be used for prediction. The input data is input into the actual short-term and short-term recurrence risk model of restricted behavior after training for risk prediction, so that the risk probability of the restricted behavior to be predicted can be obtained within the preset time range, for example, within 1 year.

Wherein, referring to FIG. 4 , the actual short-to-medium-term recurrence restriction behavior risk model may include an input layer, multiple hidden layers, and an output layer. The first layer is the input layer, and each line with each node of the next layer is a weight control parameter, denoted as w, and each node of the next layer is multiplied by each node of the previous layer by its corresponding The parameter correction of w and w can be realized through the loss function and the gradient descent algorithm, and finally a model with the best fitting effect is formed.

The input of the characteristic data to be predicted into the actual short-term re-occurrence restriction behavior risk prediction model to output the re-appearance restriction behavior prediction probability of the person to be predicted restriction activities includes:

receiving the feature data to be predicted through the input layer;

Inputting the feature data to be predicted received by the input layer into multiple hidden layers in sequence for processing;

The output layer outputs the prediction probability of the reappearance restriction behavior of the person to be predicted and obtained after the last layer of hidden layer processing.

Specifically, the input layer receives the pre-extracted feature data to be predicted of people with limited activities to be predicted, and after passing through the fully connected layer, the hidden layer1 of the first 32-dimensional node and the second hidden layer of 64 nodes Hidden layer hidden layer2, hidden layer hidden layer3 with 32 nodes, hidden layer hidden layer4 with 16 nodes, where the activation function used between each layer is the sigmod function, where x is the input value, and S(x) is the non-linear transformation The result of the specific function is as follows:

，网络输出分布为/>

，总的类别数为n：Further, the backpropagation algorithm can be used, and the loss function calculated by the corresponding error is the cross entropy function. The specific formula is as follows, where the real distribution is

, the network output distribution is />

, the total number of categories is n:

For example, if the feature vector to be predicted of a person with restricted activities is input as [1,0,1...,0,1], the corresponding real result is that there is no restricted behavior (0) again, that is, yi; the network output distribution is Refers to the prediction result obtained after the feature vector to be predicted is input into the network, for example, 0 means that it is consistent with the real result, and if it is 1, it means that it is inconsistent with the real result.

为第i个节点的输出值，C为输出节点的个数，即分类的类别个数：The 16 nodes of the last hidden layer will finally pass the Softmax function to obtain the output layer of two nodes, where the function expression of Softmax is, where

is the output value of the i-th node, and C is the number of output nodes, that is, the number of classification categories:

The output corresponding to these two nodes is a two-dimensional probability vector group, which represents the probability of the restricted behavior occurring again in one year corresponding to the multi-dimensional feature vector to be predicted. This model can fit the risk of restricting behavior occurring again within one year in real scenarios. predict. Among them, e represents a constant of nature.

Further, after going through multiple hidden layers, the last hidden layer has 16 nodes, each node can correspond to a parameter, which is respectively connected to the final two result nodes, and the parameters are corrected by the loss function and the back propagation algorithm to obtain The final reasonable parameter, for example, is finally calculated to be 0.3 for the node where the risk of restrictive behavior occurs again, and 0.7 for the node where the risk of restrictive behavior does not appear again (the sum of the two prediction results is 1), then the predicted result is that there is no recurrence of restrictive behavior risk.

In an embodiment of the present application, after outputting the predicted probability of the restricted behavior of the person to be predicted based on the actual short-term re-emergence of the restricted behavior risk prediction model, it includes:

When the predicted probability of the restricted behavior of the person to be predicted to be restricted again is greater than the preset threshold, the risk of the restricted behavior of the person to be predicted to be restricted will be visualized and displayed, and a key observation list will be sent to the management platform of the restricted activity place.

In the embodiment of this application, regular or real-time monitoring can be carried out on persons with restricted activities in places with restricted activities through the actual short-term and short-term re-emergence risk prediction model. For example, if it is higher than 0.7, it can be displayed through the visualization platform of the re-occurrence restriction behavior risk inside the restricted activity venue, so that the management personnel of the restricted activity venue can view it at any time and formulate corresponding strategies. And it can also send a key observation list to the restricted activity place management platform. For example, if the predicted probability of the restricted behavior of the person whose restricted activities will reappear is higher than 0.5, they can be used as key observers.

Further, according to the predicted probability of reoccurrence of the restricted behavior, the risk degree of the restricted activity personnel can be sorted, and key display and observation can be carried out according to the sorting.

In this embodiment of the application, by analyzing the public legal documents, the personal family information of the restricted activity personnel who entered the restricted activity place due to the restricted behavior, the social life data in the restricted activity place, etc., combined with the deep neural network, and through the specified time limit Personnel are divided into positive and negative cases, and a short-to-medium-term prediction model for the re-emergence of restricted behaviors of persons with restricted activities is constructed to predict the probability of re-emergence of restricted behaviors within the preset period, and can be combined with a visualization platform to give restricted activities Venue managers provide key watch lists to improve the overall efficiency of personnel management in restricted-activity venues. Effectively improve the accuracy and efficiency of forecasting.

It should be understood that the sequence numbers of the steps in the above embodiments do not mean the order of execution, and the execution order of each process should be determined by its functions and internal logic, and should not constitute any limitation to the implementation process of the embodiment of the present invention.

In one embodiment, a multidimensional database-based prediction system for recurrence restriction behavior is provided. The multidimensional database-based prediction system for recurrence restriction behavior is the same as the method for predicting recurrence restriction behavior based on multidimensional database in the above-mentioned embodiments. correspond. As shown in FIG. 5 , the multidimensional database-based prediction system for reoccurrence restriction behavior includes a database construction unit 10 , a data set generation unit 20 , a label generation unit 30 , a training unit 50 and a prediction unit 60 . The detailed description of each functional module is as follows:

The database construction unit 10 is used to construct a multidimensional heterogeneous database, which includes the previous legal document data, the relevant data of the previous entry into the restricted activity place, the monitoring data in the previous restricted activity place, and the current restricted behavior data of legal documents;

The data set generation unit 20 is used to perform data mining and data cleaning on the previous legal document data, the relevant data of the previous entry into the restricted activity place, the monitoring data in the previous restricted activity place and the legal document data of the current restricted behavior , to generate a sample data set, the sample data set includes a training data set and a testing data set;

The label generation unit 30 is used to divide the positive and negative examples according to the restricted behavior data of the restricted activity personnel within the preset time limit, as positive and negative example labels;

A model construction unit 40, configured to construct the original short-to-medium-term reappearance restriction behavior risk prediction model;

The training unit 50 is used to select a eigenvector group of preset dimensions in the training data set through an associated query, as input data, and input it into the original short-term recurrence restriction behavior risk prediction model for prediction processing, so as to output a recurrence The predicted probability of occurrence of restrictive behavior;

The prediction unit 60 is configured to calculate a loss function value according to the predicted probability of re-occurrence restriction behavior and the positive and negative example labels, and perform iterative training on the original medium and short-term re-occurrence restriction behavior risk prediction model according to the loss function value, Until the preset conditions are met, an actual short-to-medium-term recurrence risk prediction model of restriction behavior is generated, so as to predict the risk of restriction behavior re-emergence within the preset time range of the restricted activity personnel through the actual short-term recurrence restriction risk prediction model.

In one embodiment, the data set generating unit 20 is also used for:

Obtain the parameters of intimacy between persons with different restricted activities by performing video image analysis on the monitoring data in the previous restricted activity place;

According to the intimacy parameter, calculate the intimacy of other persons who jointly participate in the restricted behavior as the data to be predicted, and classify it into the sample data set.

In an embodiment, the label generating unit 30 is also used for:

When the restrictive behavior event occurs again within the preset time limit for the restricted activity personnel, classify the corresponding restricted activity personnel in the sample data set as a negative example;

When the person with restricted activities does not reappear the restricted behavior event within the preset time limit, the corresponding person with restricted activities in the sample data set is classified as a positive case.

In an embodiment, the prediction unit 60 is also used for:

By using a preset loss function and a backpropagation algorithm, the value of the loss function is calculated for the predicted probability of the recurrence restriction behavior and the labels of positive and negative examples.

In an embodiment, the system further includes an actual prediction unit, configured to:

Obtain the legal document data, restricted behavior data, and life data in the restricted activity place of the current restricted behavior of the person to be predicted;

Perform data mining and data cleaning on the legal document data of the current restricted behavior, restricted behavior data, and life data in the restricted activity site to generate feature data to be predicted;

The to-be-predicted feature data is input into the actual short-to-medium-term recurrence risk prediction model to output the prediction probability of the recurrence restriction behavior of the person to be predicted to restrict activities.

In an embodiment, the prediction unit 60 is also used for:

receiving the feature data to be predicted through the input layer;

Inputting the feature data to be predicted received by the input layer into multiple hidden layers in sequence for processing;

The output layer outputs the prediction probability of the reappearance restriction behavior of the person to be predicted and obtained after the last layer of hidden layer processing.

In an embodiment, the system further includes: a visual display unit, configured to:

When the predicted probability of the reoccurrence of the restricted behavior of the person to be predicted to be restricted is greater than the preset threshold, the risk of the reappearance of the restricted behavior of the person to be predicted to be restricted is displayed visually, and the key observation is sent to the management platform of the restricted activity site list.

In this embodiment of the application, by analyzing the public legal documents, the personal and family information of the persons with restricted activities who entered the restricted activity places due to restricted behaviors, the social life data in the restricted activity places, etc., combined with the deep neural network, and through the Personnel are divided into positive and negative cases, and a short-to-medium-term prediction model for the reoccurrence of restricted behavior by restricted activity personnel is constructed to predict the probability of restricted activity personnel's reoccurrence of restricted behavior within the preset period, and can be combined with a visualization platform to give restricted activities Venue management personnel provide a key watch list to improve the overall efficiency of personnel management in restricted activity venues. Effectively improve the accuracy and efficiency of forecasting.

For the specific limitations of the prediction system for the recurrence restriction behavior based on the multidimensional database, please refer to the above definition for the prediction method for the recurrence restriction behavior based on the multidimensional database, which will not be repeated here. Each module in the above system can be fully or partially realized by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, and can also be stored in the memory of the computer device in the form of software, so that the processor can call and execute the corresponding operations of the above modules.

In one embodiment, a computer device is provided. The computer device may be a terminal device, and its internal structure may be as shown in FIG. 6 . The computer device includes a processor, a memory, and a network interface connected through a system bus. Wherein, the processor of the computer device is used to provide calculation and control capabilities. The memory of the computer device includes readable storage media. The readable storage medium stores computer readable instructions. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer-readable instructions are executed by the processor, a method for predicting recurrence-limited behavior based on a multidimensional database is realized. The readable storage medium provided in this embodiment includes a non-volatile readable storage medium and a volatile readable storage medium.

In one embodiment, a computer device is provided, comprising a memory, a processor, and computer-readable instructions stored in the memory and operable on the processor, the processor executing the computer-readable The instructions are the steps of implementing the method for predicting the reoccurrence limit behavior based on the multidimensional database as described above.

In one embodiment, a readable storage medium is provided, the readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the re-appearance based on the multidimensional database as described above is realized Steps in the predictive method for limiting behavior.

Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware through computer-readable instructions, and the computer-readable instructions can be stored in a non-volatile memory When being read from a storage medium or a volatile readable storage medium, the computer-readable instructions may include the processes of the embodiments of the above-mentioned methods when executed. Wherein, any references to memory, storage, database or other media used in the various embodiments provided in the present application may include non-volatile and/or volatile memory. Nonvolatile memory can include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Chain Synchlink DRAM (SLDRAM), Direct Memory Bus Dynamic RAM (DRDRAM), etc.

Those skilled in the art can clearly understand that for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used for illustration. In practical applications, the above-mentioned functions can be assigned to different functional units, Completion of modules means that the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above.

The above-described embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: it can still carry out the foregoing embodiments Modifications to the technical solutions recorded in the examples, or equivalent replacement of some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should be included in within the protection scope of the present invention.

Claims (10)

1. A method for predicting reappearance limiting behavior based on a multidimensional database, the method comprising:

constructing a multidimensional heterogeneous database, wherein the multidimensional heterogeneous database comprises previous legal document data, related data of previous access to a restricted activity place, monitoring data in the previous restricted activity place and legal document data of current restricted behavior;

Performing data mining and data cleaning on the previous legal document data, the related data of the previous access to the restricted activity place, the monitoring data in the previous restricted activity place and the legal document data of the current restricted activity to generate a sample data set, wherein the sample data set comprises a training data set and a test data set;

dividing positive examples and negative examples according to the reappearance limiting behavior data of limiting activity personnel within a preset period to serve as positive example labels and negative example labels;

constructing an original medium-short term reappearance limiting behavior risk prediction model;

selecting a feature vector group with preset dimensionality in the training data set through associated query, and inputting the feature vector group serving as input data into the original medium-short term reappearance limiting behavior risk prediction model for prediction processing so as to output reappearance limiting behavior prediction probability;

and calculating a loss function value according to the reappearance limiting behavior prediction probability, the positive example label and the negative example label, and performing iterative training on the original medium-short term reappearance limiting behavior risk prediction model according to the loss function value until a preset condition is met, generating an actual medium-short term reappearance limiting behavior risk prediction model, so as to predict reappearance limiting behavior risk in a preset time range of limiting activity personnel through the actual medium-short term reappearance limiting behavior risk prediction model.

2. The prediction method based on reappearance limiting behavior of multidimensional database according to claim 1, wherein after constructing the multidimensional heterogeneous database, the method comprises:

video image analysis is carried out on the monitoring data in the previous limited activity place, so that affinity parameters among different limited activity personnel are obtained;

and calculating the affinity of the personnel participating in the limiting action together with other personnel according to the affinity parameter, taking the affinity as data to be predicted, and classifying the data into the sample data set.

3. The prediction method of reappearance limiting behavior based on a multidimensional database according to claim 1, wherein the positive and negative case division is performed on the limiting activity personnel corresponding to the sample data set according to reappearance limiting behavior data of the limiting activity personnel within a preset period, and the method comprises the following steps:

dividing the corresponding restricted activity personnel in the sample data set into negative examples when the restricted activity personnel generate a restricted activity event again within a preset period;

and when the restricted activity personnel do not generate the restricted action event again within the preset period, dividing the restricted activity personnel corresponding to the sample data set into positive examples.

4. The multi-dimensional database based reappearance limiting behavior prediction method according to claim 1, wherein the calculating a loss function value according to the reappearance limiting behavior prediction probability and positive and negative case labels comprises:

and calculating the loss function value for the reappearance limiting behavior prediction probability, the positive example label and the negative example label through a preset loss function and a back propagation algorithm.

5. The multi-dimensional database based reappearance limiting behavior prediction method according to claim 1, wherein after generating the actual medium-short term reappearance limiting behavior risk prediction model, the method comprises:

acquiring legal document data of current limiting behaviors of limiting personnel to be predicted, related data of entering a limiting activity place due to limiting behaviors and life data in the limiting activity place;

performing data mining and data cleaning on legal document data of the current limiting behavior, related data entering a limiting activity place due to the limiting behavior and life data in the limiting activity place to generate feature data to be predicted;

and inputting the characteristic data to be predicted into the actual medium-short term reappearance limiting behavior risk prediction model so as to output reappearance limiting behavior prediction probability of the limiting movable personnel to be predicted.

6. The method for predicting the reappearance limiting behavior based on the multidimensional database according to claim 5, wherein the actual medium-short term reappearance limiting behavior risk prediction model comprises an input layer, a plurality of hidden layers and an output layer, wherein the inputting the characteristic data to be predicted into the actual medium-short term reappearance limiting behavior risk prediction model to output the reappearance limiting behavior prediction probability of the person to be predicted to limit the activity comprises:

receiving the feature data to be predicted through the input layer;

the feature data to be predicted received by the input layer are sequentially input into a plurality of hidden layers for processing;

and outputting the reappearance limiting behavior prediction probability of the to-be-predicted limiting movable personnel obtained after the last hidden layer is processed through the output layer.

7. The multi-dimensional database-based reappearance restricting behavior prediction method of claim 6, wherein after outputting the reappearance restricting behavior prediction probability of the restricting activity personnel to be predicted according to the actual medium-short term reappearance restricting behavior risk prediction model, comprising:

and when the prediction probability of the reappearance limiting behaviors of the limiting activity personnel to be predicted is larger than a preset threshold, visually displaying the reappearance limiting behavior risk of the limiting activity personnel to be predicted, and sending a key observation list to a limiting activity place management platform.

8. A multi-dimensional database based prediction system for reappearance of constraint behavior, the system comprising:

the database construction unit is used for constructing a multidimensional heterogeneous database, and the multidimensional heterogeneous database comprises former legal document data, related data of former entering a restricted activity place, monitoring data of former entering the restricted activity place and legal document data of current restricted behavior;

the data set generation unit is used for carrying out data mining and data cleaning on the previous legal document data, the related data of the previous entering restricted activity place, the monitoring data in the previous restricted activity place and the legal document data of the current restricted action so as to generate a sample data set, wherein the sample data set comprises a training data set and a test data set;

the label generating unit is used for dividing positive examples and negative examples according to the reappearance limiting behavior data of limiting active personnel within a preset period to serve as positive example labels and negative example labels;

the model construction unit is used for constructing an original medium-short term reappearance limiting behavior risk prediction model;

the training unit is used for selecting a feature vector group with preset dimensionality in the training data set through the association query, inputting the feature vector group as input data into the original medium-short term reappearance limiting behavior risk prediction model for prediction processing so as to output reappearance limiting behavior prediction probability;

And the prediction unit is used for calculating a loss function value according to the reappearance limiting behavior prediction probability and the positive and negative labels, and carrying out iterative training on the original medium-short term reappearance limiting behavior risk prediction model according to the loss function value until the preset condition is met, and generating an actual medium-short term reappearance limiting behavior risk prediction model so as to predict reappearance limiting behavior risk in a preset time range of limiting activity personnel through the actual medium-short term reappearance limiting behavior risk prediction model.

9. A computer device comprising a memory, a processor and computer readable instructions stored in the memory and executable on the processor, wherein the processor, when executing the computer readable instructions, implements the steps of the multi-dimensional database based re-occurrence restriction behavior prediction method of any one of claims 1 to 7.

10. A readable storage medium storing computer readable instructions which, when executed by a processor, implement the steps of the multi-dimensional database based re-occurrence restriction behavior prediction method of any one of claims 1 to 7.

Images