User behavior information analysis method and system, electronic equipment and storage medium

Abstract

The invention discloses a user behavior information analysis method and system, electronic equipment and a storage medium, the user behavior information analysis system comprises a terminal, an edge end and a cloud end, and the method comprises the steps that the terminal collects current situation awareness data of a user in real time and transmits the current situation awareness data to the edge end; the edge end inputs the current situation awareness data into an LSTM model deployed in the edge end, and analyzes the current situation awareness data through the LSTM model in the edge end to obtain current behavior information of the user; wherein the LSTM model in the edge end is deployed at the edge end after the LSTM model trained in the cloud end is simplified, and model parameter updating is carried out by utilizing the cloud end to feed back parameters of the latest LSTM model in the cloud end in real time; the cloud end continuously utilizes the situation awareness data collected by the edge end to optimize and update the LSTM model in the cloud end; and the edge end feeds back the current behavior information of the user to the terminal for service processing.

Description

Technical Field

[0001] The present application relates to the technical field of user information analysis, and in particular to a user behavior information analysis method and system, electronic equipment, and storage medium. Background Art

[0002] In order to better meet user needs, increase user dependence, or increase product sales, current applications usually perform user information analysis on each user, that is, analyze user needs or preferences, and make personalized content or product recommendations to users based on the analyzed user information.

[0003] One way to analyze users is to conduct statistical analysis on their recent purchase and browsing records to find out their behavior patterns and consumption habits, thereby predicting their current needs. The second way is to use machine learning algorithms, such as support vector machines (SVMs) and decision trees, to build user behavior prediction models. These models can automatically learn rules and make predictions from a large number of users' purchases and data.

[0004] However, these two methods mainly analyze temporary user behavior data and cannot capture subtle changes and trends. In addition, user behavior habits will change constantly, so the accuracy of the analysis results is limited. In addition, the first method requires large-scale data processing, which makes it difficult to achieve real-time prediction and requires a lot of resource support. As for the second method, the model training process and the size of the model itself also require a lot of resource support. Therefore, these two methods usually need to rely on cloud platforms or backend clusters for implementation, that is, the data needs to be transferred to the cloud platform or backend cluster for analysis and processing, so the data will be centrally stored and transmitted over long distances. Therefore, there is a greater risk of data privacy leakage, and the timeliness of the analysis results is also poor. Summary of the invention

[0005] Based on the above-mentioned deficiencies of the prior art, the present application provides a user behavior information analysis method and system, electronic device, and storage medium to solve the problem that the prior art has the risk of data leakage and cannot obtain accurate analysis results in a timely manner.

[0006] In order to achieve the above objectives, this application provides the following technical solutions:

[0007] The first aspect of the present application provides a user behavior information analysis method, which is applied to a user behavior information analysis system, wherein the user behavior information analysis system includes a terminal, an edge terminal, and a cloud. The user behavior information analysis method includes:

[0008] The terminal collects the user's current situation awareness data in real time and transmits the current situation awareness data to the edge end;

[0009] The edge end inputs the current situation awareness data into the LSTM model deployed in the edge end, analyzes the current situation awareness data through the LSTM model in the edge end, and predicts the current behavior information of the user; wherein the LSTM model in the edge end pre-simplifies the LSTM model trained in the cloud end, deploys it in the edge end, and uses the parameters of the latest LSTM model in the cloud end fed back by the cloud end in real time to update the model parameters; the cloud end continuously uses the situation awareness data collected by the edge end to optimize and update the LSTM model in the cloud end;

[0010] The edge end feeds back the user's current behavior information to the terminal for business processing.

[0011] Optionally, in the above-mentioned user behavior information analysis method, the terminal collects the user's current situation awareness data in real time, including:

[0012] The terminal collects the user's current location information, current behavior information, current environment information and current business handling information in real time;

[0013] The terminal combines the current location information, the current behavior information, the current environment information and the current business handling information into a unified data set to obtain the data set of the current situation awareness data.

[0014] Optionally, in the above-mentioned user behavior information analysis method, before transmitting the current situation awareness data to the edge end, it also includes:

[0015] Performing data cleaning on the current situation awareness data;

[0016] Converting the cleaned current situation awareness data into a unified format;

[0017] Performing preliminary feature extraction on the current situation awareness data after format conversion;

[0018] The features extracted from the current situation awareness data are compressed.

[0019] Optionally, in the above user behavior information analysis method, it also includes:

[0020] The cloud receives the training sample data collected by the terminal; wherein the training sample data includes situation awareness data of multiple users at various moments in a historical time period;

[0021] The cloud pre-processes the training sample data;

[0022] The cloud initializes the parameters of the LSTM model in the cloud;

[0023] The cloud uses the training sample data to iteratively train the LSTM model in the cloud until the prediction of the output result of the LSTM model in the cloud meets the preset standard;

[0024] The cloud saves the trained LSTM model and sends the simplified local LSTM model to each edge terminal;

[0025] Each of the edge terminals receives the LSTM model sent by the cloud, and loads the LSTM model into a local environment for deployment.

[0026] Optionally, in the above user behavior information analysis method, it also includes:

[0027] The edge terminal collects the current situation awareness data;

[0028] Whenever the collection of situation awareness data meets a preset collection condition, the edge end uploads the currently collected situation awareness data to the cloud; wherein the preset collection condition is that the collection time meets the preset time, or the amount of collected data meets the preset data amount;

[0029] The cloud optimizes and updates the local LSTM model using the situation awareness data uploaded by each edge terminal at every preset time interval, or whenever the amount of situation awareness data uploaded by each edge terminal reaches a threshold, and cleans up the current situation awareness data;

[0030] The cloud sends the updated model parameters of the local LSTM model to each of the edge terminals;

[0031] Each of the edge terminals updates the local LSTM model using the model parameters sent by the cloud.

[0032] The second aspect of the present application provides a user behavior information analysis system, including:

[0033] Terminals, edge devices, and cloud;

[0034] The terminal is used to collect the user's current situation awareness data in real time and transmit the current situation awareness data to the edge end;

[0035] The edge end is used to input the current situation awareness data into the LSTM model deployed in the edge end, analyze the current situation awareness data through the LSTM model in the edge end, predict the current behavior information of the user, and feed back the current behavior information of the user to the terminal for business processing, and use the parameters of the latest LSTM model in the cloud fed back by the cloud in real time to update the model parameters;

[0036] The cloud end is used to pre-train the LSTM model and deploy it to the edge end after simplifying it, and continuously use the situational awareness data collected by the edge end to optimize and update the LSTM model on the cloud end, and feed back the latest parameters of the LSTM model to the edge end.

[0037] Optionally, in the above-mentioned user behavior information analysis system, when the terminal collects the user's current situation awareness data in real time, it is used to:

[0038] Collect the user's current location information, current behavior information, current environment information, and current business handling information in real time;

[0039] The current location information, the current behavior information, the current environment information and the current business handling information are combined into a unified data set to obtain a data set of the current situation awareness data.

[0040] Optionally, in the above user behavior information analysis system, the terminal is further used for:

[0041] Performing data cleaning on the current situation awareness data;

[0042] Converting the cleaned current situation awareness data into a unified format;

[0043] Performing preliminary feature extraction on the current situation awareness data after format conversion;

[0044] The features extracted from the current situation awareness data are compressed.

[0045] Optionally, in the above-mentioned user behavior information analysis system, the LSTM model is pre-trained in the cloud and simplified, and then deployed to the edge end for:

[0046] Receiving training sample data collected by the terminal; wherein the training sample data includes situation awareness data of multiple users at various moments in a historical time period;

[0047] Preprocessing the training sample data;

[0048] Initializing the parameters of the LSTM model in the cloud;

[0049] The cloud uses the training sample data to iteratively train the LSTM model in the cloud until the prediction of the output result of the LSTM model in the cloud meets the preset standard;

[0050] The trained LSTM model is saved, and the simplified local LSTM model is sent to each edge terminal to trigger each edge terminal to receive the LSTM model sent by the cloud, and load the LSTM model into the local environment for deployment.

[0051] Optionally, in the above-mentioned user behavior information analysis system, when the cloud continues to optimize and update the LSTM model of the cloud using the situation awareness data collected by the edge end, and feeds back the latest parameters of the LSTM model to the edge end, it is used to:

[0052] At every preset time interval, or whenever the amount of the situation awareness data uploaded by each edge terminal reaches a threshold, the local LSTM model is optimized and updated using the situation awareness data uploaded by each edge terminal, and the current situation awareness data is cleaned up; wherein the situation awareness data is collected by the edge terminal, and whenever the collection of the situation awareness data meets a preset collection condition, the currently collected situation awareness data is uploaded to the cloud; the preset collection condition is that the collection time meets the preset time, or the amount of data collected meets the preset data amount;

[0053] The updated model parameters of the local LSTM model are sent to each of the edge terminals to trigger each of the edge terminals to update the local LSTM model using the model parameters sent by the cloud.

[0054] A third aspect of the present application provides an electronic device, including:

[0055] Memory and processor;

[0056] Wherein, the memory is used to store programs;

[0057] The processor is used to execute the program, and when the program is executed, it is specifically used to implement the user behavior information analysis method as described in any one of the above.

[0058] A fourth aspect of the present application provides a computer storage medium for storing a computer program, which, when executed by a processor, is used to implement a user behavior information analysis method as described in any one of the above.

[0059] The present application provides a method for analyzing user behavior information, which is implemented by a user behavior information analysis system. Among them, the user behavior information analysis system includes a terminal, an edge terminal and a cloud. When performing user behavior information analysis, the terminal collects the user's current situation awareness data in real time, and transmits the current situation awareness data to the edge terminal, so that more subtle information can be captured through the current situation awareness data analysis, and the accuracy of the result is provided. Then the edge terminal inputs the current situation awareness data into the LSTM model deployed in the edge terminal, analyzes the current situation awareness data through the LSTM model in the edge terminal, predicts the user's current behavior information, and feeds back the user's current behavior information to the terminal for business processing. Thus, user behavior information analysis is realized through the LSTM model in the edge terminal. Since there is no need to analyze and concentrate a large amount of data, and the edge terminal is close to the terminal, the risk of data leakage can be reduced, and the timeliness of data analysis can be provided. The LSTM model in the edge terminal simplifies the LSTM model trained in the cloud in advance and deploys it on the edge terminal, so that it is trained through the cloud model with more resources, and the model is simplified and provided to the edge terminal for use, so that the edge terminal with limited resources has the ability to accurately perform data analysis. In addition, the cloud side continuously uses the situational awareness data collected by the edge side to optimize and update and feed back to the edge side, so that the edge side can use the latest LSTM model parameters fed back by the cloud side in real time to update the model parameters, ensuring that the LSTM model on the edge side adapts to user changes in real time and maintains the function of accurately analyzing data. Thus, through the combined application of terminals, edge sides and cloud sides, a method for accurately analyzing user behavior information in a safe and timely manner is realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] 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 embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0061] Figure 1 A flowchart of a user behavior information analysis method provided in an embodiment of the present application;

[0062] Figure 2 A flowchart of a method for a terminal to collect a user's current situation awareness data in real time provided in an embodiment of the present application;

[0063] Figure 3 A flowchart of a method for preprocessing current situation awareness data provided in an embodiment of the present application;

[0064] Figure 4A flowchart of a training method for an LSTM model provided in an embodiment of the present application;

[0065] Figure 5 A flowchart of an optimization and updating method of an LSTM model provided in an embodiment of the present application;

[0066] Figure 6 A schematic diagram of the architecture of a user behavior information analysis system provided in an embodiment of the present application;

[0067] Figure 7 A schematic diagram of the architecture of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0068] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0069] In this application, relational terms such as first and second, etc. are used only 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 "comprise", "include" 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, an element defined by the statement "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device including the element.

[0070] The present application embodiment provides a user behavior information analysis method, which is applied to a user behavior information analysis system. The user behavior information analysis system includes a terminal, an edge terminal, and a cloud. Figure 1 As shown, a user behavior information analysis method provided in an embodiment of the present application includes the following steps:

[0071] S101. The terminal collects the user's current situation awareness data in real time and transmits the current situation awareness data to the edge end.

[0072] Specifically, when any terminal needs to analyze the user's behavior information, it can collect the user's current situation awareness data. However, it should be noted that before collecting the user's current situation awareness data, the user's authorization is required.

[0073] In order to capture subtle changes and trends in user behavior and provide accuracy of analysis results, in an embodiment of the present application, situational awareness data is collected for analysis, which includes multiple types of perceived situation data, including environmental information, behavioral information, location information, etc.

[0074] In order to quickly analyze the data, obtain the analysis results and feed them back to the terminal for content recommendation, in the embodiment of the present application, real-time analysis of user behavior information is performed through the edge end. Therefore, when the terminal collects the user's current situation awareness data, it will send the current situation awareness data to the edge end closest to it for processing.

[0075] Optionally, in another embodiment of the present application, a specific implementation method of the terminal collecting the user's current situation awareness data in real time is as follows: Figure 2 As shown, the following steps are included:

[0076] S201. The terminal collects the user's current location information, current behavior information, current environment information, and current business handling information in real time.

[0077] Optionally, the user's current location information can be obtained in real time through technologies such as GPS, Wi-Fi positioning or base station positioning. Current behavior information may include information on each user's operation behavior, such as clicks, slides, and dwell time. Current environmental information includes perception information such as temperature, humidity, and light, which can be obtained through the weather application in the terminal. Current business processing information is information related to business processing, that is, usually historical business processing information up to the current moment, such as the user's purchase history, browsing history, and other consumption-related data.

[0078] S202. The terminal combines the current location information, current behavior information, current environment information and current business handling information into a unified data set to obtain a data set of current situation awareness data.

[0079] Optionally, in order to ensure the quality of collected data, quickly transmit data, and allow the edge end to quickly process data, in another embodiment of the present application, the terminal needs to further pre-process the current situation awareness data before transmitting the current situation awareness data to the edge end. Figure 3 As shown, a method for preprocessing current situation awareness data provided by an embodiment of the present application includes the following steps:

[0080] S301. Clean the current situation awareness data.

[0081] Optionally, data cleaning may specifically include removing duplicate, invalid or erroneous data records, filling missing values, and smoothing noisy data to reduce interference with subsequent data analysis.

[0082] S302: Convert the cleaned current situation awareness data into a unified format.

[0083] Since different data have different formats and cannot be subsequently processed, in an embodiment of the present application, data from different sources and in different formats are uniformly converted into a format that the system can process, such as converting timestamps into standard time formats, converting text data into numerical data, etc.

[0084] S303: Perform preliminary feature extraction on the current situation awareness data after format conversion.

[0085] It should be noted that since the data is not needed for direct analysis, but statistics or extracted features are needed for analysis, the terminal first extracts features from the current situational awareness data, such as statistical features, time series features, etc., so that the edge can quickly perform higher-level processing.

[0086] S304: compress the features extracted from the current situation awareness data.

[0087] In order to reduce transmission costs and improve processing efficiency, the features extracted from the current situational awareness data need to be compressed before transmission.

[0088] S102: The edge end inputs the current situation awareness data into the LSTM model deployed in the edge end, and the current situation awareness data is analyzed by the LSTM model in the edge end to predict the user's current behavior information.

[0089] In order to predict the user's behavior information, in the embodiment of the present application, an LSTM model is used for analysis, which can capture the long-term medical relationship in the user's behavior, thereby more accurately predicting the user's future behavior.

[0090] Considering the problem that the training of the model and the size of the model require a lot of resources, and the resources of the edge are limited. Therefore, in the embodiment of the present application, the LSTM model is pre-trained in the cloud, that is, the LSTM model is trained using sufficient resources of cloud users. Since the edge only needs to use the LSTM model for prediction and does not need to consider other issues of the model, the LSTM model trained in the cloud is simplified to retain the function of analyzing data for prediction, and then deployed to each edge, thereby effectively reducing the LSTM model and reducing the demand for resources without affecting the accuracy of its analysis results.

[0091] Moreover, in an embodiment of the present application, considering that the consumption habits of users may change differently, in order to ensure that the LSTM model can adapt to the changes in the habits of users and can always accurately predict user behavior information, it is necessary to continuously optimize the LSTM model in the edge. The LSTM model in the edge is a simplified model, and the resources of the edge are limited. Therefore, in an embodiment of the present application, the situational awareness data collected by the edge is continuously used by the cloud to optimize and update the LSTM model in the cloud. The model parameters of the latest LSTM model after optimization and update are then fed back to the edge, so that the edge can use the cloud to feedback the parameters of the latest LSTM model in the cloud in real time, and update the model parameters of the LSTM model in the edge.

[0092] Optionally, another embodiment of the present application provides a training method for an LSTM model, such as Figure 4 As shown, the following steps are included:

[0093] S401, the cloud receives training sample data collected by the terminal.

[0094] The training sample data includes situation awareness data of multiple users at various moments in a historical time period.

[0095] It should be noted that, since time series data needs to be analyzed, it is necessary to obtain situation awareness data at each moment in the historical time period.

[0096] S402: The cloud pre-processes the training sample data.

[0097] For more information, please refer to the preprocessing method for training sample data. Figure 3 The method shown will not be repeated here.

[0098] S403: The cloud initializes the parameters of the LSTM model in the cloud.

[0099] Optionally, the parameters of the LSTM model, such as weights and bias parameters, may be randomly initialized.

[0100] S404: The cloud uses the training sample data to iteratively train the LSTM model in the cloud until the prediction of the output result of the LSTM model in the cloud meets the preset standard.

[0101] Specifically, the input data is fed into the LSTM model, and after being processed for multiple time steps, the model outputs a prediction of the user's future behavior information. At each time step, the LSTM memory unit captures and retains information that is useful for subsequent predictions. The error between the model's output and the actual user behavior data is then calculated using the loss function. Based on the calculated error loss, the parameters of the LSTM model are updated using the back propagation algorithm. In addition, during the training process, the cloud also uses a validation dataset to evaluate the performance of the model, prevent overfitting, and ensure that the model can perform well on unseen data.

[0102] S405: The trained LSTM model is saved in the cloud, and the simplified local LSTM model is sent to each edge terminal.

[0103] Specifically, the simplified LSTM model can be a model part or a distilled version of the model.

[0104] S406. Each edge terminal receives the LSTM model sent from the cloud, and loads the LSTM model into the local environment for deployment.

[0105] Optionally, in another embodiment of the present application, an optimization update method of an LSTM model is provided, such as Figure 5 As shown, including:

[0106] S501, the edge end collects current situation awareness data.

[0107] It should be noted that in the embodiment of the present application, after the edge end receives the current situation awareness data, it will process it and store the current situation awareness data so as to facilitate updating the model.

[0108] S502. Whenever the collection of situation awareness data meets the preset collection conditions, the edge end uploads the currently collected situation awareness data to the cloud.

[0109] The preset collection condition is that the collection time meets the preset time, or the collected data volume meets the preset data volume.

[0110] In order to learn the latest information of users in time when updating the model, the collected data needs to be uploaded to the cloud regularly to update the model, and the collection time is reset at this time. When the amount of data is large, it means that the user has a lot of recent data, which may change, and is sufficient to optimize and update the model. Therefore, when the amount of collected data meets the preset data amount, the recently collected data will also be uploaded to the cloud.

[0111] S503. At every preset time interval on the cloud, or whenever the amount of situational awareness data uploaded by each edge terminal reaches a threshold, the situational awareness data uploaded by each edge terminal is used to optimize and update the local LSTM model, and the current situational awareness data is cleaned up.

[0112] S504: The cloud sends the updated model parameters of the local LSTM model to each edge terminal.

[0113] S505. Each edge terminal updates the local LSTM model using the model parameters sent by the cloud.

[0114] S103: The edge end feeds back the user's current behavior information to the terminal for business processing.

[0115] Specifically, the edge feeds back the user's current behavior information to the terminal, and the terminal can use the user's current behavior information to match the corresponding products or content and display them on the front-end page to provide them to the user.

[0116] The embodiment of the present application provides a method for analyzing user behavior information, which is implemented by a user behavior information analysis system. Among them, the user behavior information analysis system includes a terminal, an edge terminal and a cloud. When performing user behavior information analysis, the terminal collects the user's current situation awareness data in real time, and transmits the current situation awareness data to the edge terminal, so that more subtle information can be captured through the current situation awareness data analysis, and the accuracy of the result is provided. Then the edge terminal inputs the current situation awareness data into the LSTM model deployed in the edge terminal, analyzes the current situation awareness data through the LSTM model in the edge terminal, predicts the user's current behavior information, and feeds back the user's current behavior information to the terminal for business processing. Thus, the user behavior information analysis is realized through the LSTM model in the edge terminal, because there is no need to analyze and concentrate a large amount of data, and the edge terminal is close to the terminal, so the risk of data leakage can be reduced, and the timeliness of data analysis can be provided. The LSTM model in the edge terminal simplifies the LSTM model trained in the cloud in advance and deploys it on the edge terminal, so that it is trained through the cloud model with more resources, and the model is simplified and provided to the edge terminal for use, so that the edge terminal with limited resources has the ability to accurately perform data analysis. In addition, the cloud side continuously uses the situational awareness data collected by the edge side to optimize and update and feed back to the edge side, so that the edge side can use the latest LSTM model parameters fed back by the cloud side in real time to update the model parameters, ensuring that the LSTM model on the edge side adapts to user changes in real time and maintains the function of accurately analyzing data. Thus, through the combined application of terminals, edge sides and cloud sides, a method for accurately analyzing user behavior information in a safe and timely manner is realized.

[0117] Another embodiment of the present application provides a user behavior information analysis system, such as Figure 6 As shown, including:

[0118] Terminal 601, edge terminal 602 and cloud 603.

[0119] Among them, the terminal 601 is used to collect the user's current situation awareness data in real time and transmit the current situation awareness data to the edge terminal 602.

[0120] The edge end 602 is used to input the current situational awareness data into the LSTM model deployed in the edge end 602, analyze the current situational awareness data through the LSTM model in the edge end 602, predict the user's current behavior information, and feed back the user's current behavior information to the terminal 601 for business processing, and use the cloud 603 to feed back the latest LSTM model parameters in the cloud in real time to update the model parameters.

[0121] The cloud 603 is used to pre-train the LSTM model and deploy it to the edge 602 after simplifying it, and continuously use the situational awareness data collected by the edge 602 to optimize and update the LSTM model of the cloud 603, and feed back the latest LSTM model parameters to the edge 502.

[0122] Optionally, in the user behavior information analysis system provided in another embodiment of the present application, when the terminal collects the user's current situation awareness data in real time, it is used to:

[0123] Collect the user's current location information, current behavior information, current environment information and current business processing information in real time.

[0124] The current location information, current behavior information, current environment information and current business handling information are combined into a unified data set to obtain a data set of current situational awareness data.

[0125] Optionally, in the user behavior information analysis system provided in another embodiment of the present application, the terminal is further used for:

[0126] Perform data cleansing on current situational awareness data.

[0127] Convert the cleaned current situation awareness data into a unified format.

[0128] Perform preliminary feature extraction on the current situation awareness data after format conversion.

[0129] Compress the features extracted from the current situation awareness data.

[0130] Optionally, in a user behavior information analysis system provided in another embodiment of the present application, after the LSTM model is pre-trained and simplified in the cloud, when deployed to the edge end, it is used to:

[0131] The training sample data collected by the receiving terminal includes situation awareness data of multiple users at various moments in a historical time period.

[0132] Preprocess the training sample data.

[0133] Initialize the parameters of the LSTM model in the cloud.

[0134] The cloud uses the training sample data to iteratively train the LSTM model in the cloud until the prediction of the output result of the LSTM model in the cloud meets the preset standard.

[0135] Save the trained LSTM model and send the simplified local LSTM model to each edge end to trigger each edge end to receive the LSTM model sent by the cloud and load the LSTM model into the local environment for deployment.

[0136] Optionally, in the user behavior information analysis system provided in another embodiment of the present application, the cloud continues to optimize and update the LSTM model of the cloud using the situational awareness data collected by the edge end, and feeds back the latest parameters of the LSTM model to the edge end, for:

[0137] At every preset time interval, or whenever the amount of situational awareness data uploaded by each edge terminal reaches a threshold, the situational awareness data uploaded by each edge terminal is used to optimize and update the local LSTM model, and the current situational awareness data is cleaned up. The situational awareness data is collected by the edge terminal, and whenever the collection of situational awareness data meets the preset collection conditions, the currently collected situational awareness data is uploaded to the cloud. The preset collection conditions are that the collection time meets the preset time, or the amount of data collected meets the preset data amount.

[0138] The updated model parameters of the local LSTM model are sent to each edge terminal to trigger each edge terminal to update the local LSTM model using the model parameters sent by the cloud.

[0139] It should be noted that the specific working processes of the terminal, edge terminal and cloud terminal in the system provided in the above embodiment can refer to the implementation process of the corresponding steps in the above method embodiment, and will not be repeated here.

[0140] Another embodiment of the present application provides an electronic device, such as Figure 7 As shown, including:

[0141] Memory 701 and processor 702 .

[0142] The memory 701 is used to store programs.

[0143] The processor 702 is used to execute the program stored in the memory 701. When the program is executed, it is specifically used to implement the user behavior information analysis method provided in any one of the above embodiments.

[0144] Another embodiment of the present application provides a computer storage medium for storing a computer program, which, when executed by a processor, is used to implement a user behavior information analysis method as provided in any of the above embodiments.

[0145] Computer storage media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include temporary computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0146] 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.

[0147] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those 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 application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A user behavior information analysis method, characterized in that: Applied to a user behavior information analysis system, the user behavior information analysis system includes a terminal, an edge terminal and a cloud, and the user behavior information analysis method includes: The terminal collects the user's current situation awareness data in real time and transmits the current situation awareness data to the edge end; The edge end inputs the current situation awareness data into the LSTM model deployed in the edge end, analyzes the current situation awareness data through the LSTM model in the edge end, and predicts the current behavior information of the user; wherein the LSTM model in the edge end pre-simplifies the LSTM model trained in the cloud end, deploys it in the edge end, and uses the parameters of the latest LSTM model in the cloud end fed back by the cloud end in real time to update the model parameters; the cloud end continuously uses the situation awareness data collected by the edge end to optimize and update the LSTM model in the cloud end; The edge end feeds back the user's current behavior information to the terminal for business processing.

2. The method according to claim 1, characterized in that The terminal collects the user's current situation awareness data in real time, including: The terminal collects the user's current location information, current behavior information, current environment information and current business handling information in real time; The terminal combines the current location information, the current behavior information, the current environment information and the current business handling information into a unified data set to obtain the data set of the current situation awareness data.

3. The method according to claim 1, characterized in that Before transmitting the current situation awareness data to the edge terminal, the method further includes: Performing data cleaning on the current situation awareness data; Converting the cleaned current situation awareness data into a unified format; Performing preliminary feature extraction on the current situation awareness data after format conversion; The features extracted from the current situation awareness data are compressed.

4. The method according to claim 1, characterized in that Also includes: The cloud receives the training sample data collected by the terminal; wherein the training sample data includes situation awareness data of multiple users at various moments in a historical time period; The cloud pre-processes the training sample data; The cloud initializes the parameters of the LSTM model in the cloud; The cloud uses the training sample data to iteratively train the LSTM model in the cloud until the prediction of the output result of the LSTM model in the cloud meets the preset standard; The cloud saves the trained LSTM model and sends the simplified local LSTM model to each edge terminal; Each of the edge terminals receives the LSTM model sent by the cloud, and loads the LSTM model into a local environment for deployment.

5. The method according to claim 1, characterized in that Also includes: The edge terminal collects the current situation awareness data; Whenever the collection of situation awareness data meets a preset collection condition, the edge end uploads the currently collected situation awareness data to the cloud; wherein the preset collection condition is that the collection time meets the preset time, or the amount of collected data meets the preset data amount; The cloud optimizes and updates the local LSTM model using the situation awareness data uploaded by each edge terminal at every preset time interval, or whenever the amount of situation awareness data uploaded by each edge terminal reaches a threshold, and cleans up the current situation awareness data; The cloud sends the updated model parameters of the local LSTM model to each of the edge terminals; Each of the edge terminals updates the local LSTM model using the model parameters sent by the cloud.

6. A user behavior information analysis system, characterized in that: include: Terminals, edge devices, and cloud; The terminal is used to collect the user's current situation awareness data in real time and transmit the current situation awareness data to the edge end; The edge end is used to input the current situation awareness data into the LSTM model deployed in the edge end, analyze the current situation awareness data through the LSTM model in the edge end, predict the current behavior information of the user, and feed back the current behavior information of the user to the terminal for business processing, and use the parameters of the latest LSTM model in the cloud fed back by the cloud in real time to update the model parameters; The cloud end is used to pre-train the LSTM model and deploy it to the edge end after simplifying it, and continuously use the situational awareness data collected by the edge end to optimize and update the LSTM model on the cloud end, and feed back the latest parameters of the LSTM model to the edge end.

7. The system according to claim 6, characterized in that When the terminal collects the user's current situation awareness data in real time, it is used to: Collect the user's current location information, current behavior information, current environment information, and current business handling information in real time; The current location information, the current behavior information, the current environment information and the current business handling information are combined into a unified data set to obtain a data set of the current situation awareness data.

8. The system according to claim 6, characterized in that The terminal is further used for: Performing data cleaning on the current situation awareness data; Converting the cleaned current situation awareness data into a unified format; Performing preliminary feature extraction on the current situation awareness data after format conversion; The features extracted from the current situation awareness data are compressed.

9. An electronic device, characterized in that: include: Memory and processor; Wherein, the memory is used to store programs; The processor is used to execute the program, and when the program is executed, it is specifically used to implement the user behavior information analysis method as described in any one of claims 1 to 5.

10. A computer storage medium, characterized in that: Used to store a computer program, which, when executed by a processor, is used to implement the user behavior information analysis method as described in any one of claims 1 to 5.

Images