User behavior analysis method and device, computer equipment and storage medium

Abstract

The invention relates to a user behavior analysis method and device, computer equipment and a storage medium. The method comprises the steps of obtaining user behavior data, including order operation data and resource operation data, of a user for public cloud service within a preset time period, determining an order operation index of the user based on the order operation data, determining a resource operation index of the user based on the resource operation data, a user behavior index is determined according to the order operation data index and the resource operation data index, the user behavior index serves as input, a preset user activity level prediction model is called, the user activity level of the user is predicted, and the user activity level prediction model is composed of a historical user behavior index carrying a user activity level label and a historical user behavior index carrying a user activity level label. And training a preset recurrent neural network model. By adopting the method, the accuracy of a user behavior analysis result can be improved.

Description

Technical Field

[0001] The present application relates to the field of big data technology, and in particular to a user behavior analysis method, apparatus, computer equipment and storage medium. Background Art

[0002] With the development of big data technology and the acceleration of social digital transformation, the cloud computing industry has assumed an increasingly important mission as a digital infrastructure. The application scope of public cloud is becoming wider and wider, and the number of users is also increasing. Therefore, how to analyze the operational behavior of public cloud users has become one of the problems to be solved.

[0003] In traditional solutions, users' operation behaviors and usage experiences of public cloud services are usually collected through questionnaires and user-initiated feedback, or users' operation behaviors are analyzed by searching the public cloud's operation logs.

[0004] However, the above-mentioned method based on user feedback may lead to user subjectivity in the user behavior analysis results, and the number of user samples and the screening scope are limited, making it difficult to fully reflect user behavior. In addition, since the data in the operation log is messy and scattered, and there is a lot of dirty data, the analysis method based on the operation log will lead to low accuracy of user behavior analysis results. Summary of the invention

[0005] Based on this, it is necessary to provide a user behavior analysis method, apparatus, computer equipment, computer-readable storage medium and computer program product that can improve the accuracy of user behavior analysis in response to the above technical problems.

[0006] In a first aspect, the present application provides a user behavior analysis method. The method comprises:

[0007] Obtaining user behavior data of users for public cloud services within a preset time period, wherein the user behavior data includes order operation data and resource operation data;

[0008] Determining an order operation index of the user based on the order operation data;

[0009] Determining a resource operation indicator of the user based on the resource operation data;

[0010] Determining a user behavior indicator according to the order operation data indicator and the resource operation data indicator;

[0011] Taking the user behavior indicator as input, calling a preset user activity level prediction model to predict the user activity level of the user;

[0012] The user activity level prediction model is obtained by training a preset recurrent neural network model based on historical user behavior indicators that carry user activity level labels.

[0013] In a second aspect, the present application also provides a user behavior analysis device. The device comprises:

[0014] A data acquisition module, used to acquire user behavior data of users for public cloud services within a preset time period, wherein the user behavior data includes order operation data and resource operation data;

[0015] An order operation data processing module, used to determine the order operation index of the user based on the order operation data;

[0016] A resource operation data processing module, used to determine the resource operation index of the user based on the resource operation data;

[0017] A user behavior processing module, used to determine a user behavior indicator according to the order operation data indicator and the resource operation data indicator;

[0018] An activity prediction module, used to use the user behavior indicator as input, call a preset user activity level prediction model, and determine the user activity level of the user;

[0019] The user activity level prediction model is obtained by training a preset recurrent neural network model based on historical user behavior indicators that carry user activity level labels.

[0020] In a third aspect, the present application further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps in the above user behavior analysis method embodiment are implemented.

[0021] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps in the above-mentioned user behavior analysis method embodiment are implemented.

[0022] In a fifth aspect, the present application further provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps in the above user behavior analysis method embodiment are implemented.

[0023] The above-mentioned user behavior analysis method, device, computer equipment and storage medium are different from the traditional user feedback and operation log analysis methods. By directly extracting the user behavior data of the user within a preset time period, including order operation data and resource operation data, it can more objectively reflect the user's actual operation behavior, and then determine the user's order operation indicators based on the order operation data, and determine the user's resource operation indicators based on the resource operation data. Finally, the user behavior indicators are determined according to the order operation data indicators and the resource operation data indicators, so that the user behavior data can be converted into quantifiable indicator data, thereby more accurately describing the user's behavior characteristics. Finally, the user activity level prediction model is called to analyze the user's user activity level. Since the user activity level prediction model is based on the historical user behavior indicators carrying the user activity level label, and is obtained by training the preset recurrent neural network model, the recurrent neural network model can learn to identify complex time dependencies during the training process. Therefore, the trained user activity level prediction model can accurately identify the relationship between the user behavior indicators and the user activity level, thereby improving the accuracy of user behavior analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 An application environment diagram of a user behavior analysis method in one embodiment;

[0025] Figure 2 A flowchart of a user behavior analysis method in one embodiment;

[0026] Figure 3 A flowchart of steps for determining a user behavior indicator in one embodiment;

[0027] Figure 4 A flowchart of steps for determining user behavior indicators in another embodiment;

[0028] Figure 5 is a flowchart of a user behavior analysis method in another embodiment;

[0029] Figure 6 A flowchart of a user behavior analysis method in a detailed embodiment;

[0030] Figure 7 is a structural block diagram of a user behavior analysis device in one embodiment;

[0031] Figure 8 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0033] The user behavior analysis method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown, the terminal 102 communicates with the server 104 through a network. The data storage system can store data that the server 104 needs to process. The data storage system can be integrated on the server 104, or placed on the cloud or other network servers.

[0034] Specifically, the analyst may upload the user behavior data of the user within a preset time period to the server 104 through the terminal 102, or the server 104 may directly retrieve the user behavior data, including the order operation data and the resource operation data, from the data storage system. Then, the server 104 determines the user's order operation index based on the order operation data, determines the user's resource operation index based on the resource operation data, determines the user behavior index based on the order operation data index and the resource operation data index, and uses the user behavior index as input to call the preset user activity level prediction model to predict the user's user activity level. Among them, the user activity level prediction model can be obtained by pre-training a preset recurrent neural network model with historical user behavior indicators carrying user activity level labels, and the trained user activity level prediction model can be pre-stored in the data storage system of the server 104.

[0035] The terminal 102 may be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. The portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers.

[0036] In one embodiment, Figure 2 As shown, a user behavior analysis method is provided, which is applied to Figure 1 Taking the server 104 in the example as an example, the following steps are included:

[0037] S100, obtaining user behavior data of users for public cloud services within a preset time period.

[0038] Among them, user behavior data includes order operation data and resource operation data. The server can be the server of the public cloud console. User behavior data refers to the data generated when the user performs order operations or resource operations in the public cloud console. The public cloud console processes its behavior data and stores it in mongodb.

[0039] For example, when a user orders, allocates, renews, cancels, or retrieves a cloud service product through the public cloud console, and places an order through the order management layer of the public cloud console, the order operation data corresponding to the above order operation is stored in the database of the public cloud console backend, including but not limited to the order number, order creation time, order type, resource type, resource specification, order status, user ID, resource pool ID, etc., where the order number is a unique ID. For another example, when a user performs some free resource operations through the public cloud console, such as powering on, shutting down, restarting, and reinstalling a cloud host, and binding a cloud resource to a disk or network card, the resource operation data generated includes but is not limited to the cloud resource ID, operation time, operation type, operation status, user ID, resource pool ID, etc.

[0040] When analysts wish to retrieve the above user behavior data, they can pre-set a collection cycle, such as defining the collection of user behavior data within the past month, and then call the console backend interface through the front end of the public cloud console to extract the user behavior data generated in the preset time period from the backend database.

[0041] S200, determining the order operation index of the user based on the order operation data, and determining the resource operation index of the user based on the resource operation data.

[0042] Among them, the order operation index is the quantification of the user's order operation for the public cloud service, and the resource operation index is the quantification of the user's resource operation for the public cloud service. The order operation data and resource operation data can be cleaned and counted respectively, and the order operation index obtained includes but is not limited to the total order operation amount, order operation time interval, etc., and the resource operation index obtained includes but is not limited to the total resource operation amount, resource operation time interval, etc.

[0043] S300, determining a user behavior indicator according to the order operation data indicator and the resource operation data indicator.

[0044] Among them, user behavior indicators refer to quantitative indicators used to describe user behavior obtained by analyzing, summarizing, extracting features from user behavior data, including but not limited to user login frequency, single usage duration, number of operation types, operation time intervals, etc.

[0045] For example, the order operation indicators and resource operation indicators may be further counted and summarized to analyze and obtain the total number of user operations on the public cloud console (for example, the number of user operations in a day), the average operation time interval (for example, in a day, the user operates once every 1 hour on average), etc.

[0046] S400, taking the user behavior index as input, calling a preset user activity level prediction model to predict the user's user activity level.

[0047] The user activity level prediction model is obtained by training a preset recurrent neural network model with historical user behavior indicators carrying user activity level labels. The recurrent neural network model is a neural network structure used to process time series data, which can learn the time dependency in time series data. For example, after training with historical user behavior indicators carrying user activity level labels, the user activity level prediction model obtained can classify users into high-activity and low-activity according to user behavior indicators.

[0048] Exemplarily, the server first uses historical user behavior indicators with user activity level labels, including but not limited to historical user behavior indicators of a large number of users in different historical time periods, and corresponding activity labels, to train a recurrent neural network model to obtain a user activity level prediction model. At this time, user behavior analysis is performed on user A, which can be input into the user activity level prediction model. The user activity level prediction model extracts the characteristics of the user behavior indicators, classifies them, and outputs the user activity level of user A, which can be "highly active" or "lowly active".

[0049] The above-mentioned user behavior analysis method is different from the traditional user feedback and operation log analysis methods. It directly extracts the user behavior data of the user within a preset time period, including order operation data and resource operation data, so it can more objectively reflect the user's actual operation behavior, and then determine the user's order operation indicators based on the order operation data, and determine the user's resource operation indicators based on the resource operation data. Finally, the user behavior indicators are determined according to the order operation data indicators and the resource operation data indicators, so that the user behavior data can be converted into quantifiable indicator data, thereby more accurately describing the user's behavior characteristics. Finally, the user activity level prediction model is called to analyze the user's user activity level. Since the user activity level prediction model is based on the historical user behavior indicators carrying the user activity level label, and is obtained by training the preset recurrent neural network model, the recurrent neural network model can learn to recognize complex time dependencies during the training process. Therefore, the trained user activity level prediction model can accurately identify the relationship between user behavior indicators and user activity levels, thereby improving the accuracy of user behavior analysis.

[0050] In one embodiment, Figure 3 As shown, S200 includes:

[0051] S210, determining the total amount of order operations of the user within a preset time period and the time interval of the order operations according to the order operation data.

[0052] S220: Determine the total amount of resource operations performed by the user within a preset time period and the resource operation time interval according to the resource operation data.

[0053] Among them, order operation data refers to the operations related to billing resources performed by users in the public cloud console, such as ordering, changing configurations, renewing, and unsubscribing billing cloud resources. The total number of order operations refers to the cumulative number of order operations performed by users within a preset time period (the unit can be orders / day), which is used to characterize the frequency of interaction between users and the public cloud console. The order operation time interval refers to the time interval between two consecutive order operations performed by users within a preset time period. The median or average of all time intervals can be taken to characterize the user's activity in order operations.

[0054] Resource operation data refers to the operations performed by users on free cloud resources in the public cloud console. Taking cloud hosts as an example, free cloud resource operations include but are not limited to power on, power off, restart, reinstall, cloud resource binding and other operations. The total number of resource operations is the number of times users perform resource operations within a preset time period. The resource operation time interval is the time interval between users performing resource operations within the preset time period, which can be expressed as the median or average of all time intervals.

[0055] Exemplarily, the user behavior data of all users in one day can be filtered out from the backend database of the public cloud console using a mongodb query statement, and then the user behavior data can be grouped and summed according to the user identifier to obtain the total number of order operations (orders / day) and total number of resource operations (times / day) of each user, and then the time intervals of all order operations and all resource operations of each user in one day are counted, and the order operation time interval (order / time) and resource operation time interval (hour / time) are obtained by taking the average value. Exemplarily, if according to the order operation data, it is counted that user A has performed 5 order operations in one day, the total number of order operations is 5 orders / day, and the time intervals of these 5 order operations are 10 minutes, 15 minutes, 5 minutes, and 10 minutes respectively, then the order operation time interval is calculated to be 10 minutes / order by taking the average value. According to the above method, the user behavior data of the preset time period is processed.

[0056] In this embodiment, the total number of order operations, the time interval of order operations, the total number of resource operations, and the time interval of resource operations can well reflect the user's operating habits. Therefore, by analyzing the above-mentioned order operation indicators and resource operation indicators, the user's operating habits on the public cloud console can be quickly quantified, so as to facilitate the subsequent efficient and accurate extraction of user behavior characteristics and analysis of user activity levels.

[0057] In one embodiment, Figure 3 As shown, S300 includes:

[0058] S310, determining the total amount of user behavior according to the total amount of order operations and the total amount of resource operations.

[0059] S320, determining the user behavior time interval according to the order operation time interval and the resource operation time interval.

[0060] S330, determining a user behavior index according to the total amount of user behavior and the time interval of user behavior.

[0061] The total amount of user behavior refers to the cumulative number of all user operations within a preset time period, including the sum of the total order operations and the total resource operations. The user behavior time interval is a comprehensive reflection of the order operation time interval and the resource operation time interval within the preset time period, and can be the average or weighted average of the above two time intervals. User behavior indicators are standardized features obtained based on the total amount of user behavior and the user behavior time interval, which can reflect the user's user activity level and behavior pattern.

[0062] Exemplarily, the total amount of user behavior = the total amount of order operations + the total amount of resource operations, and the user behavior time interval = (order operation time interval + resource operation time interval) / 2. Based on the total amount of user behavior and the user behavior time interval, the user behavior index can be obtained by processing based on a preset standardization method, for example, the total amount of user behavior and the user behavior time interval are normalized, and the normalized result is expressed in the form of an array to obtain the user behavior index.

[0063] In this embodiment, the total amount of user behavior and the time interval of user behavior can well reflect the user's operating habits. Therefore, by counting the total amount of user behavior and the time interval of user behavior, a reliable basis can be provided for the subsequent efficient and accurate extraction of user behavior characteristics and analysis of user activity levels.

[0064] In one embodiment, Figure 4 As shown, S330 includes:

[0065] S331, normalize the total amount of user behaviors and the time interval of user behaviors to obtain a normalized result.

[0066] S332: construct a user behavior array based on the normalization result, where the user behavior array represents the user behavior within a preset time period.

[0067] S333, determining a user behavior index according to the user behavior array.

[0068] Among them, normalization is a combing method that can convert data values ​​of different dimensions into a specific range, thereby eliminating the order of magnitude differences between different data indicators. User behavior array is a data structure that can be used to store user behavior characteristics within a preset time period. User behavior indicators can be features or statistics further extracted based on the user behavior array. In addition to characterizing the total amount of user behavior and the time interval of user behavior, they can also characterize characteristics such as user operation frequency.

[0069] For example, the preset time period can be 30 days, which is divided into three time periods, each of which is 10 days. First, the total amount of user behavior and the time interval of user behavior are mapped to the range of [0, 1] by linear normalization. The specific normalization formula can be shown as formula (1):

[0070] (1)

[0071] In formula (1), y is the normalized result, x is the total amount of user behavior or the time interval of user behavior, Indicates the minimum value of the total amount of user behavior or the time interval of user behavior. Indicates the total amount of user behavior or the maximum value in a time interval of user behavior.

[0072] Then, with 10 days as a time period, the normalized results are constructed into a three-dimensional array through numpy data in python. The size of the array is (m, 10, 2), where m is the size of the sample size. In this embodiment, the sample size refers to the number of samples of the total amount of user behavior or the time interval of user behavior within a preset time period. 10 means that a time period is 10 days, and 2 means that a user has two behavioral characteristics, which refers to the total amount of user behavior and the time interval of user behavior in this embodiment. In this way, the three-dimensional array can be determined as a user behavior array, and the normalized results are stored in the user behavior array according to time. Therefore, based on the user behavior array, user behavior indicators with time series characteristics can be determined.

[0073] In this embodiment, the numerical differences between data of different dimensions can be reduced through normalization processing, and the user behavior characteristics can be abstracted into structured data by constructing a user behavior array. The time series characteristics of the user behavior array can not only reflect the changing trend of user behavior, but also provide rich time series characteristics for the activity prediction model, so that subsequent user behavior analysis can be performed more efficiently and accurately.

[0074] In one embodiment, Figure 5 As shown, after S400, the method further includes:

[0075] S510 , when the user activity level is the first activity level, determine the user's behavior category, the operation frequency corresponding to the behavior category, and the operation success rate corresponding to the behavior category according to the user behavior data.

[0076] S520, determining a target behavior category in which the operation frequency is greater than a preset frequency threshold and the operation success rate is less than a preset success rate threshold.

[0077] S530: Determine, based on the user behavior data, the failure reason why the operation success rate of the target behavior category is less than a preset success rate threshold.

[0078] S610: When the user activity level is the second activity level, a user churn risk reminder message is pushed.

[0079] Among them, the first activity level is higher than the second activity level. User behavior categories refer to different types of operations performed by users on the public cloud console, such as order creation, cloud host activation, cloud resource application, cloud resource allocation, etc. Operation frequency refers to the frequency of user operations in a certain behavior category, and operation success rate refers to the proportion of successful operations in a certain behavior category. The first type of active users refers to highly active users, and more detailed user behavior analysis is needed for highly active users to enhance user stickiness. The second type of active users refers to low-activity users. Since low-activity users have a higher risk of churn, user churn risk reminder messages need to be pushed.

[0080] For example, when the user activity level is the first activity level, that is, the user is a highly active user, the user's behavior category, the operation frequency corresponding to the behavior category, and the operation success rate corresponding to the behavior category are counted according to the user behavior data. For example, the order process such as the activation and configuration change of the cloud host, the operation process such as the power on and off, etc., the calculation formula of the operation success rate can be shown as formula (2):

[0081] (2)

[0082] In formula (2), Indicates the success rate of the operation. represents the number of successful user behaviors in the i-th behavior category, Represents the total amount of user behaviors in the i-th behavior category. Further, from the behavior categories, the target behavior categories with operation frequency greater than the preset frequency threshold and operation success rate less than the preset success rate threshold are screened out, that is, the behavior categories that frequently appear in the user behavior data but have a relatively low operation success rate. For the target behavior category, the failure reasons for the relatively low operation success rate of this behavior category can be analyzed manually or by deep learning methods, including but not limited to system errors, user input errors, network instability, etc., and then the products in the public cloud console can be optimized to improve user satisfaction.

[0083] In addition, for highly active users, we can further count the cloud resource specifications and resource pools that have been activated more frequently by users (greater than the preset number threshold) to conduct user behavior preference analysis, thereby launching targeted marketing activities. For example, we can launch some promotional activities for cloud hosts with specifications corresponding to resource pools that have been activated more frequently by users, thereby further improving user satisfaction.

[0084] When the user activity level is the second level, that is, the user is a low-activity user, the public cloud console can automatically push a user churn risk reminder message to the front end for low-activity users, where the user churn risk reminder message can carry a low-activity user identifier, indicating that these users have low login frequency and few operations on the public cloud console. After the front-end staff receives the user churn risk reminder message through the front end, they can send coupons, product usage questionnaires, and return visits to the user accounts of low-activity users to understand the user's experience with the public cloud console and encourage users to return.

[0085] In this embodiment, for highly active users, by analyzing their behavior categories, operation frequencies, and operation success rates, it is possible to identify the behavior categories of highly active users with high operation frequencies but low operation success rates, thereby optimizing the products in the public cloud console in a targeted manner, such as optimizing the operation process, improving the user interface, etc., to improve the satisfaction of highly active users. For low-activity users, by pushing user churn risk reminder messages, the staff is instructed to adopt some incentive mechanisms, thereby reducing the user churn rate and improving the user retention rate of the public cloud console.

[0086] In one embodiment, the user activity level prediction model includes a gated recurrent unit and a classification layer connected in sequence. When the user activity level prediction model is called, the following steps are performed: taking the user behavior indicator as input, calling the gated recurrent unit, extracting the user behavior time series feature data, taking the user behavior time series feature data as input, calling the classification layer, and determining the user activity level.

[0087] Among them, the user behavior index characterizes the time series statistical characteristics of the user's operation behavior on the public cloud platform. The gated recurrent unit is a recurrent neural network structure that is suitable for processing time series data and can capture the time dependency of data in multiple time steps. The user behavior time series feature data can be extracted through the gating mechanism. In this embodiment, a three-layer gated recurrent unit can be used for feature extraction to improve the prediction effect. The classification layer classifies users according to the user behavior time series feature data, and divides users into high-activity users and low-activity users. The classification layer in this embodiment can be a softmax classifier of the fully connected layer.

[0088] For example, when the user activity level prediction model is called, the user behavior index is first input into the gated recurrent unit, which can capture the temporal behavior pattern of the user behavior and output the user behavior time series feature data. It should be noted that the gated recurrent unit includes a reset gate and an update gate, and the calculation formulas of each gate and memory cell are shown in formulas (3) to (6):

[0089] (3)

[0090] (4)

[0091] (5)

[0092] (6)

[0093] In the above formula, is the sigmoid function; To reset the gate; To update the door; is the current candidate hidden state; , , Represent the weight matrices of the reset gate, update gate, and candidate hidden states respectively; Represents the current hidden layer state; Both represent vector concatenation; Both represent bias terms.

[0094] Furthermore, the user behavior time series feature data is passed to the classification layer. Through the softmax function, the classification layer can map users to different activity categories based on the user behavior time series feature data, such as labeling users with high activity or low activity. , then the sample The probability of being classified into category j is calculated as shown in formula (7):

[0095] (7)

[0096] In this embodiment, by extracting user behavior time series feature data through a gated recurrent unit, the trend of user activity changing over time can be effectively captured, thereby identifying the long-term and short-term characteristics of user activity level changes, and providing more effective information for the subsequent classification layer. The classification layer then classifies users based on the user behavior time series feature data to obtain highly accurate activity classification results.

[0097] In one embodiment, a user activity level prediction model is trained based on the following steps: obtaining historical user behavior indicators, labeling the historical user behavior indicators, obtaining historical user behavior indicators carrying user activity level labels, and training a pre-trained recurrent neural network model based on the historical user behavior indicators carrying the user activity level labels to obtain a user activity level prediction model, wherein the pre-trained recurrent neural network model includes a gated recurrent unit and a classification layer connected in sequence.

[0098] Among them, the historical user behavior indicators are past user behavior data. The historical user behavior indicators can be divided into a training set and a test set. The data in the training set is used in the training process of the model so that the model can learn the characteristics of different activity categories, and the data in the test set is used to test the performance of the trained model. The pre-trained recurrent neural network model means that the recurrent neural network model has been preliminarily trained on some general data and has a certain classification ability. It only needs to be trained with a small amount of specific data (in this embodiment, historical user behavior indicators with user activity level labels) to obtain a trained model.

[0099] The process of labeling historical user behavior data is the process of classifying or labeling users according to historical user behavior indicators. For example, based on the historical user behavior indicators of user A, it is determined that user A is a low-activity user, and the historical user behavior indicators of user A are labeled with a user activity level label of "low activity". Furthermore, the historical user behavior indicators carrying the user activity level label are divided into a training set and a test set. Based on the training set, the preset recurrent neural network model is trained and tuned for multiple rounds to obtain multiple sets of model parameters. Finally, the model parameters with the best performance on the test set are selected as the target model parameters to construct a user activity level prediction model.

[0100] In this embodiment, based on historical user behavior indicators carrying user activity level labels, the recurrent neural network model can gradually learn the different behavioral characteristic patterns of high-activity and low-activity users. The trained user activity level prediction model can identify user behavior time series feature data and has strong generalization ability, and can effectively and accurately predict the activity categories of different users.

[0101] In order to make a clearer description of the user behavior analysis method provided in this application, Figure 6 and one A detailed embodiment is provided to explain the user behavior analysis method, and the detailed embodiment includes the following steps:

[0102] S601, obtaining order operation data and resource operation data of a user for a public cloud service within a preset time period.

[0103] S602, determining the total amount of order operations and the time interval of order operations of the user within a preset time period based on the order operation data, and determining the total amount of resource operations and the time interval of resource operations of the user within a preset time period based on the resource operation data.

[0104] S603, determining the total amount of user behavior according to the total amount of order operations and the total amount of resource operations, and determining the user behavior time interval according to the order operation time interval and the resource operation time interval.

[0105] S604, normalize the total amount of user behaviors and the time interval of user behaviors to obtain a normalized result.

[0106] S605 , constructing a user behavior array according to the normalization result, where the user behavior array represents the user behavior within a preset time period.

[0107] S606, determining a user behavior index according to the user behavior array, taking the user behavior index as input, calling a preset user activity level prediction model, and determining the user activity level of the user.

[0108] S607 : When the user activity level is the first activity level, determine the user's behavior category, the operation frequency corresponding to the behavior category, and the operation success rate corresponding to the behavior category according to the user behavior data.

[0109] S608, determining a target behavior category whose operation frequency is greater than a preset frequency threshold and whose operation success rate is less than a preset success rate threshold, and determining the failure reason why the operation success rate of the target behavior category is less than the preset success rate threshold based on the user behavior data.

[0110] S609: When the user activity level is the second activity level, a user churn risk reminder message is pushed.

[0111] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0112] Based on the same inventive concept, the embodiment of the present application also provides a user behavior analysis device for implementing the user behavior analysis method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more user behavior analysis device embodiments provided below can refer to the limitations on the user behavior analysis method above, and will not be repeated here.

[0113] In one embodiment, Figure 7 As shown, a user behavior analysis device 700 is provided, comprising: a data acquisition module 710, an order operation data processing module 720, a resource operation data processing module 730, a user behavior processing module 740 and an activity prediction module 750, wherein:

[0114] The data acquisition module 710 is used to acquire user behavior data of users for public cloud services within a preset time period, where the user behavior data includes order operation data and resource operation data.

[0115] The order operation data processing module 720 is used to determine the user's order operation index based on the order operation data.

[0116] The resource operation data processing module 730 is used to determine the user's resource operation index based on the resource operation data.

[0117] The user behavior processing module 740 is used to determine the user behavior index according to the order operation data index and the resource operation data index.

[0118] The activity prediction module 750 is used to use the user behavior index as input, call a preset user activity level prediction model, and determine the user activity level of the user.

[0119] Among them, the user activity level prediction model is obtained by training a preset recurrent neural network model based on historical user behavior indicators carrying user activity level labels.

[0120] In one embodiment, the order operation data processing module 720 is further used to determine the total amount of order operations of the user within a preset time period and the time interval of the order operations according to the order operation data, and the order operation index includes the total amount of order operations and the time interval of the order operation. The resource operation data processing module 730 is further used to determine the total amount of resource operations of the user within a preset time period and the time interval of the resource operations according to the resource operation data, and the resource operation index includes the total amount of resource operations and the time interval of the resource operations.

[0121] In one embodiment, the user behavior processing module 740 is also used to determine the total amount of user behavior based on the total amount of order operations and the total amount of resource operations, determine the user behavior time interval based on the order operation time interval and the resource operation time interval, and determine the user behavior indicator based on the total amount of user behavior and the user behavior time interval.

[0122] In one embodiment, the user behavior processing module 740 is also used to normalize the total amount of user behavior and the time interval of user behavior to obtain a normalized result. Based on the normalized result, a user behavior array is constructed. The user behavior array represents the user behavior within a preset time period. Based on the user behavior array, the user behavior index is determined.

[0123] In one embodiment, the user behavior analysis device 700 is also used to determine the user's behavior category, the operation frequency corresponding to the behavior category, and the operation success rate corresponding to the behavior category based on the user behavior data when the user activity level is the first activity level, determine the target behavior category with an operation frequency greater than a preset frequency threshold and an operation success rate less than a preset success rate threshold, determine the failure reason for the operation success rate of the target behavior category being less than the preset success rate threshold based on the user behavior data, and push a user churn risk reminder message when the user activity level is the second activity level, wherein the activity level of the first active type users is higher than the activity level of the second active type users.

[0124] In one embodiment, the user activity level prediction model includes a gated recurrent unit and a classification layer connected in sequence. The activity prediction module 730 is also used to take the user behavior indicator as input, call the gated recurrent unit, extract the user behavior time series feature data, take the user behavior time series feature data as input, call the classification layer, and determine the user activity level.

[0125] In one embodiment, the user behavior analysis device 700 is also used to obtain historical user behavior indicators, label the historical user behavior indicators, obtain historical user behavior indicators carrying user activity level labels, and train a pre-trained recurrent neural network model based on the historical user behavior indicators carrying the user activity level labels to obtain a user activity level prediction model. The pre-trained recurrent neural network model includes a gated recurrent unit and a classification layer connected in sequence.

[0126] Each module in the above user behavior analysis device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0127] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 8 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data such as user behavior data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a user behavior analysis method is implemented.

[0128] Those skilled in the art will understand that Figure 8 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0129] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the steps in the above-mentioned user behavior analysis method embodiment are implemented.

[0130] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned user behavior analysis method embodiment are implemented.

[0131] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above-mentioned user behavior analysis method embodiment when executed by a processor.

[0132] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0133] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0134] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0135] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the present application. It should be noted that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A user behavior analysis method, characterized in that: The method comprises: Obtaining user behavior data of users for public cloud services within a preset time period, wherein the user behavior data includes order operation data and resource operation data; Determining an order operation index of the user based on the order operation data; Determining a resource operation indicator of the user based on the resource operation data; Determining a user behavior indicator according to the order operation data indicator and the resource operation data indicator; Taking the user behavior indicator as input, calling a preset user activity level prediction model to predict the user activity level of the user; The user activity level prediction model is obtained by training a preset recurrent neural network model based on historical user behavior indicators that carry user activity level labels.

2. The method according to claim 1, characterized in that The determining the order operation index of the user based on the order operation data includes: Determine the total amount of order operations and the time interval of order operations of the user within a preset time period according to the order operation data, wherein the order operation index includes the total amount of order operations and the time interval of order operations; The determining the resource operation index of the user based on the resource operation data includes: According to the resource operation data, the total amount of resource operations and the resource operation time interval of the user within a preset time period are determined, and the resource operation indicator includes the total amount of resource operations and the resource operation time interval.

3. The method according to claim 2, characterized in that Determining the user behavior indicator according to the order operation data indicator and the resource operation data indicator includes: Determine the total amount of user behavior according to the total amount of order operations and the total amount of resource operations; Determine the user behavior time interval according to the order operation time interval and the resource operation time interval; A user behavior index is determined based on the total amount of user behavior and the user behavior time interval.

4. The method according to claim 3, characterized in that The determining of the user behavior index based on the total amount of user behavior and the time interval of the user behavior includes: Normalizing the total amount of user behavior and the time interval of the user behavior to obtain a normalized result; Constructing a user behavior array according to the normalization result, wherein the user behavior array represents the user behavior within a preset time period; A user behavior index is determined according to the user behavior array.

5. The method according to any one of claims 1 to 4, characterized in that: After predicting the user activity level of the user, the method further includes: When the user activity level is the first activity level, determining, according to the user behavior data, a behavior category of the user, an operation frequency corresponding to the behavior category, and an operation success rate corresponding to the behavior category; Determine a target behavior category in which the operation frequency is greater than a preset frequency threshold and the operation success rate is less than a preset success rate threshold; Determine, based on the user behavior data, a failure reason why the operation success rate of the target behavior category is less than a preset success rate threshold; When the user activity level is the second activity level, pushing a user churn risk reminder message; The first activity level is higher than the second activity level.

6. The method according to any one of claims 1 to 4, characterized in that: The user activity level prediction model includes a gated recurrent unit and a classification layer connected in sequence. When the user activity level prediction model is called, the following steps are performed: Taking the user behavior indicator as input, calling the gated recurrent unit, and extracting the user behavior time series feature data; The classification layer is called with the user behavior time series feature data as input to determine the user activity level.

7. The method according to any one of claims 1 to 4, characterized in that: The user activity level prediction model is trained based on the following steps: Get historical user behavior metrics; Labeling the historical user behavior indicators to obtain historical user behavior indicators carrying user activity level labels; According to the historical user behavior indicators carrying the user activity level labels, a pre-trained recurrent neural network model is trained to obtain the user activity level prediction model, wherein the pre-trained recurrent neural network model includes a gated recurrent unit and a classification layer connected in sequence.

8. A user behavior analysis device, characterized in that: The device comprises: A data acquisition module, used to acquire user behavior data of users for public cloud services within a preset time period, wherein the user behavior data includes order operation data and resource operation data; An order operation data processing module, used to determine the order operation index of the user based on the order operation data; A resource operation data processing module, used to determine the resource operation index of the user based on the resource operation data; A user behavior processing module, used to determine a user behavior indicator according to the order operation data indicator and the resource operation data indicator; An activity prediction module, used to use the user behavior indicator as input, call a preset user activity level prediction model, and determine the user activity level of the user; The user activity level prediction model is obtained by training a preset recurrent neural network model based on historical user behavior indicators that carry user activity level labels.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images