Interaction qualification level determination method, device, equipment and storage medium

Abstract

The present application discloses a method, apparatus, device and storage medium for determining an interaction qualification level, the method comprising: obtaining a first prediction intermediate result of the first-stage behavior data of a target user, performing a longitudinal federation prediction with the second device based on the first prediction intermediate result to obtain an aggregated result of the first stage; determining the first-stage interaction qualification level of the target user based on the aggregated result of the first stage to determine the second-stage behavior data; performing a longitudinal federation prediction with the second device based on the second-stage behavior data to obtain the second-stage aggregated result of the target user to obtain the second-stage interaction qualification level of the target user, wherein the second stage is a time-series continuous stage with the first stage. The present application aims to solve the technical problem of low accuracy in determining user interaction qualification levels in the prior art.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology in financial technology (Fintech), and in particular to a method, device, equipment and storage medium for determining an interaction qualification level. Background Art

[0002] With the continuous development of financial technology, especially Internet technology finance, more and more technologies are applied in the financial field, but the financial industry also puts forward higher requirements for technology. For example, the financial industry also has higher requirements for the determination of interactive qualification levels.

[0003] At present, the user's interaction qualification level is usually determined by the order rate prediction results determined by the behavior data distributed in a specific dimension. However, in most scenarios, the feature richness of the data distributed in a specific dimension is low. If the user's interaction qualification level is determined in the above way, the user's order rate prediction results will be inaccurate, which will lead to inaccurate determination of the user's interaction qualification level. Summary of the invention

[0004] The main purpose of the present application is to provide a method, device, equipment and storage medium for determining an interaction qualification level, aiming to solve the technical problem in the prior art that it is difficult to accurately determine the interaction qualification level of a user.

[0005] To achieve the above object, the present application provides a method for determining an interaction qualification level, the method comprising:

[0006] Obtaining a first prediction intermediate result of the first-stage behavior data of the target user, and performing a vertical federated prediction with the second device based on the first prediction intermediate result to obtain an aggregated result of the first stage;

[0007] Based on the aggregation results of the first stage, determining the first stage interaction qualification level of the target user to determine the second stage behavior data;

[0008] Based on the second-stage behavior data, a longitudinal federated prediction is performed with the second device to obtain the second-stage aggregated result of the target user to obtain the second-stage interaction qualification level of the target user, wherein the second stage is a temporally continuous stage with the first stage.

[0009] Optionally, the step of obtaining a first prediction intermediate result of the first-stage behavior data of the target user includes:

[0010] Acquire the first-stage behavior data of the target user, and determine the first-stage characteristics of the first-stage behavior data;

[0011] A first target prediction model of the first stage features is obtained, and based on the first target prediction model, a first order rate prediction is performed on the target user to obtain a first prediction intermediate result.

[0012] Optionally, the step of obtaining the first target prediction model of the first stage features includes:

[0013] Acquire first device training data with preset labels of the first stage features, and acquire second device training data of the second device in the first stage;

[0014] A preset basic model is obtained, and based on the first device training data with preset labels and the second device training data, the preset basic model is federatedly trained until a preset termination training condition is reached to obtain the first target prediction model.

[0015] Optionally, the step of performing vertical federated prediction with the second device based on the first prediction intermediate result to obtain an aggregated result of the first stage includes:

[0016] Receiving a second prediction intermediate result of the second device behavior data of the target user in the first stage determined by the second device;

[0017] The first prediction intermediate result and the second prediction intermediate result are aggregated to obtain an aggregated result of the first stage.

[0018] Optionally, the step of determining the first-stage interaction qualification level of the target user based on the aggregation result of the first stage to determine the second-stage behavior data includes:

[0019] Based on the aggregation results of the first stage, determining the first stage interaction qualification level of the target user;

[0020] determining a first interaction file based on the first-stage interaction qualification level;

[0021] The first interaction file is parsed to determine the second-stage behavior data.

[0022] Optionally, the step of parsing the first interaction file to determine the second-stage behavior data includes:

[0023] Determine the channel type of the target user;

[0024] The first interaction file is parsed, and second-stage behavior data is determined based on the channel type and the parsed first interaction file.

[0025] Optionally, after the step of performing longitudinal federated prediction with the second device based on the second-stage behavior data to obtain the second-stage aggregated result of the target user to obtain the second-stage interaction qualification level of the target user, the method further comprises:

[0026] If it is detected that the order rate prediction results of the target user in multiple time-series continuous stages are all greater than the preset value, a preset special identification processing is performed on the target user;

[0027] A preset targeted recommendation process is executed for the target user after the preset special identification processing.

[0028] Optionally, the step of determining the first-stage interaction qualification level of the target user based on the aggregation result of the first stage to determine the second-stage behavior data includes:

[0029] Determining the first-stage interaction qualification level of the first-stage target user based on the aggregation result of the first stage;

[0030] Obtain the channel type of the first-stage interaction qualification level;

[0031] If the channel type is a public account channel, determine to output a preset selected article to the target user based on the first level and the public account channel;

[0032] The second-stage behavior data of the target user determined based on the preset selected articles is obtained.

[0033] The present application also provides an interactive qualification level determination device, the interactive qualification level determination device comprising:

[0034] A first acquisition module, used to acquire a first prediction intermediate result of the first-stage behavior data of the target user, and perform a vertical federated prediction with the second device based on the first prediction intermediate result to obtain an aggregated result of the first stage;

[0035] A determination module, used to determine the first-stage interaction qualification level of the target user based on the aggregation result of the first stage, so as to determine the second-stage behavior data;

[0036] The second acquisition module is used to perform longitudinal federated prediction with the second device based on the second-stage behavior data to obtain the second-stage aggregation result of the target user to obtain the second-stage interaction qualification level of the target user, wherein the second stage is a time-series continuous stage with the first stage.

[0037] Optionally, the first acquisition module includes:

[0038] A first acquisition unit, configured to acquire first-stage behavior data of a target user and determine first-stage features of the first-stage behavior data;

[0039] The second acquisition unit is used to acquire a first target prediction model of the first stage characteristics, and based on the first target prediction model, perform a first order rate prediction on the target user to obtain a first prediction intermediate result.

[0040] Optionally, the second acquiring unit includes:

[0041] A first acquisition subunit is used to acquire first device training data with preset labels of the first stage features, and acquire second device training data of the second device in the first stage;

[0042] The second acquisition subunit is used to acquire a preset basic model, and based on the first device training data with preset labels and the second device training data, perform federated iterative training on the preset basic model until a preset termination training condition is reached to obtain the first target prediction model.

[0043] Optionally, the first acquisition module includes:

[0044] A receiving unit, configured to receive a second prediction intermediate result of the second device behavior data of the target user in the first stage determined by the second device;

[0045] An aggregation unit is used to aggregate the first prediction intermediate result and the second prediction intermediate result to obtain an aggregation result of the first stage.

[0046] Optionally, the determining module includes:

[0047] A third acquisition unit, configured to acquire the aggregation result based on the first stage, and determine the first stage interaction qualification level of the target user;

[0048] A first determining unit, configured to determine a first interaction file based on the first-stage interaction qualification level;

[0049] The second determining unit is used to parse the first interaction file and determine the second stage behavior data.

[0050] Optionally, the second determining unit includes:

[0051] A first determining subunit, configured to determine a channel type of the target user;

[0052] The second determining subunit is configured to parse the first interaction file and determine the second-stage behavior data based on the channel type and the parsed first interaction file.

[0053] Optionally, the interaction qualification level determination device further includes:

[0054] A detection module, configured to perform a preset special identification process on the target user if it is detected that the order rate prediction results of the target user in multiple time-series continuous stages are all greater than a preset value;

[0055] The execution module is used to execute a preset targeted recommendation process for the target user after the preset special identification processing.

[0056] Optionally, the determining module includes:

[0057] A third determining unit, configured to determine a first-stage interaction qualification level of a first-stage target user based on the aggregation result of the first stage;

[0058] A fourth obtaining unit, configured to obtain the channel type of the first-stage interaction qualification level;

[0059] A fourth determining unit, configured to determine, if the channel type is a public account channel, to output a preset selected article to the target user based on the first level and the public account channel;

[0060] The fifth acquisition unit is used to acquire the second-stage behavior data of the target user determined based on the preset selected articles.

[0061] The present application also provides an interaction qualification level determination device, which is a physical device, and includes: a memory, a processor, and a program of the interaction qualification level determination method stored in the memory and executable on the processor. When the program of the interaction qualification level determination method is executed by the processor, the steps of the interaction qualification level determination method as described above can be implemented.

[0062] The present application also provides a storage medium, on which is stored a program for implementing the above-mentioned interaction qualification level determination method. When the program of the interaction qualification level determination method is executed by a processor, the steps of the above-mentioned interaction qualification level determination method are implemented.

[0063] The present application obtains the first prediction intermediate result of the first-stage behavior data of the target user, and performs vertical federation prediction with the second device based on the first prediction intermediate result to obtain the first-stage aggregation result; based on the first-stage aggregation result, the first-stage interaction qualification level of the target user is determined to determine the second-stage behavior data; based on the second-stage behavior data, the vertical federation prediction is performed with the second device to obtain the second-stage aggregation result of the target user to obtain the second-stage interaction qualification level of the target user, wherein the second stage and the first stage are time-series continuous stages. Compared with the prior art that determines the user's interaction qualification level by determining the order rate prediction result of the behavior data distributed in a specific dimension, the present application can achieve the user order rate prediction result of the corresponding stage by stage or grade, and then obtain the target user's interaction qualification level, and in the present application, by vertical federation with other devices, the richness of the data is broadened, so that the order rate is predicted based on richer behavior data, thereby improving the prediction accuracy of the order rate prediction result, and then improving the determination accuracy of the target user's interaction qualification level, overcoming the technical defect of low accuracy of determination of the target user's interaction qualification level in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0065] 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, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0066] Figure 1 This is a flow chart of the first embodiment of the interactive qualification level determination method of the present application;

[0067] Figure 2 This is a flowchart of detailed steps of step S20 in the interactive qualification level determination method of this application;

[0068] Figure 3 This is a schematic diagram of the device structure of the hardware operating environment involved in the embodiment of the present application.

[0069] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0070] 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.

[0071] The present application embodiment provides a method for determining an interaction qualification level. In the first embodiment of the method for determining an interaction qualification level of the present application, refer to Figure 1 , the interaction qualification level determination method includes:

[0072] Step S10, obtaining a first prediction intermediate result of the first-stage behavior data of the target user, and performing a vertical federated prediction with the second device based on the first prediction intermediate result to obtain an aggregated result of the first stage;

[0073] Step S20, based on the aggregation result of the first stage, determining the first stage interaction qualification level of the target user to determine the second stage behavior data;

[0074] Step S30 performs vertical federated prediction with the second device based on the second-stage behavior data to obtain the second-stage aggregation result of the target user to obtain the second-stage interaction qualification level of the target user, wherein the second stage is a time-series continuous stage with the first stage.

[0075] The specific steps are as follows:

[0076] Step S10, obtaining a first prediction intermediate result of the first-stage behavior data of the target user, and performing a vertical federated prediction with the second device based on the first prediction intermediate result to obtain an aggregated result of the first stage;

[0077] In this embodiment, it should be noted that it can be applied to an interactive qualification level determination system, which is subordinate to an interactive qualification level determination device. For the interactive qualification level determination system, it has a communication connection relationship with an organization such as an enterprise. Therefore, as long as the user has an operation behavior on the enterprise's associated web page, the interactive qualification level determination system can obtain the user's operation behavior, and based on the user's operation behavior, predict the user's order possibility or purchase possibility and the level of the order possibility, and then determine the target user's interactive qualification level. Then the enterprise provides feedback service based on the order possibility or purchase possibility. The feedback service can be carried out through various channels such as public accounts, WeChat, APP, etc., among which the feedback service channel is associated with the channel corresponding to the user's operation behavior. For the enterprise, no matter what channel is used, there are industry characteristics (business scope characteristics), and the industry Features include business features, and business features include different product features, etc. That is, therefore, for enterprises, multi-layer and deep serial and parallel path services are provided for user operations. It should be noted that the enterprise's feedback service or path service is different for different users' different operating behaviors on the enterprise website. Although enterprises can interact with users through apps, mini-programs, official accounts or sharing links, and enterprises provide multi-layer and deep serial and parallel path services, overall, the user's interaction with the enterprise includes serial paths such as enterprise, business, portal, and page. After the serial path, there is a parallel path, such as: article path, problem path, dialogue path, contact path, such as adding WeChat, making a phone call, etc. Among them, the serial path is an operating behavior path that all users have, while the parallel path is not available to all users, or for a certain user at a certain stage, there may be only one parallel path.

[0078] Specifically, an enterprise and a user (user end) can interact through multiple entrances or channels, such as through an app, through a mini-program, through a public account, or through a sharing link. After the enterprise has recorded the operation behavior data on the app, mini-program, public account, or sharing link, for example, after the enterprise obtains the operation behavior data of each user at different depths through the operation data records of each operation button, the operation behavior data of each user at different depths is sent to the interaction qualification level determination system to predict whether the user will purchase or place an order. In this embodiment, it should be noted that the operational behavior data recorded by the enterprise includes at least one user behavior. For example, the user clicks to read an article, adds customer service on WeChat, and answers a customer service call. Then, each user behavior is clicked to read the article, added on WeChat, and answered a call, and each user behavior is phased. The phased nature here may refer to: after the first user behavior, the user receives services from the enterprise based on the first user behavior feedback, such as the feedback on the WeChat public account article (if the first user behavior is different, the WeChat public account article will be different), and subsequently generates user behaviors (such as telephone behaviors) based on the feedback services, such as the feedback on the WeChat public account article. Then, the behaviors generated thereafter are different from the first user behavior stage. In this embodiment, predictions are made based on the data corresponding to the changed behaviors. After obtaining the corresponding prediction results, the enterprise can dynamically adjust the service level and the corresponding services based on the level of the prediction results.

[0079] Overall, for example, on the stage timeline, first obtain the portrait data of user A, then add the user WeChat to recommend item B to user A, and then recommend item B to user A over the phone. Among them, the first stage behavior data is the portrait data provided by the user, for example, the user fills in his or her hobbies, etc. After the first stage, the second stage is the WeChat communication stage, and the user's second stage behavior data is the WeChat communication text data. The third stage is the telephone communication stage (if the second stage is set as the first stage at the start, the third stage is the second stage), and the user's third stage behavior data is the telephone communication voice data

[0080] The first prediction intermediate result of the first-stage behavior data of the target user is obtained. Specifically, the first-stage behavior data of the target user is obtained, and the first target prediction model of the first-stage behavior data is determined to obtain the first prediction intermediate result; wherein the first target prediction model is obtained by performing a federated prediction with the second device in the first stage. The second device has the same user as the first device, but the first prediction task corresponding to the first-stage behavior data is inconsistent with the second prediction task corresponding to the second device in the first stage. For example, if the user intends to take out a loan to buy an insurance product, the first prediction task may be to predict the probability of the user buying the insurance product, and the second prediction task may be to predict the probability of the user taking out a loan.

[0081] The step of obtaining the first prediction intermediate result of the first stage behavior data of the target user includes:

[0082] Step S11, obtaining the first-stage behavior data of the target user, and determining the first-stage characteristics of the first-stage behavior data;

[0083] Step S12, obtaining a first target prediction model of the first stage features, and based on the first target prediction model, performing a first order rate prediction on the target user to obtain a first prediction intermediate result.

[0084] Acquire the first-stage behavior data of the target user, and determine the first-stage features of the first-stage behavior data. The first-stage features may include only features x1, x2, x3 and x4, or only features x1, x2, x3, x4 and x5, or only features x1, x2, x3, x4 and x6, or only features x1, x2, x3, x4, x5 and x7, or only features x1, x2, x3, x4, x5, x6 and x7, etc.

[0085] The method of obtaining the first-stage behavior data of the target user and determining the first-stage characteristics of the first-stage behavior data includes:

[0086] Method 1:

[0087] After obtaining the first-stage behavior data of the target user, the serial-parallel path of the first-stage behavior data is obtained through the first-stage behavior data, and the first-stage characteristics of the first-stage behavior data are determined through the serial-parallel path of the first-stage behavior data.

[0088] Method 2: After obtaining the first-stage behavior data of the target user, obtain the path channel, path level, path depth, etc. of the first-stage behavior data through the first-stage behavior data, and determine the first-stage characteristics of the first-stage behavior data through the path channel, path level, path depth, etc.

[0089] Specifically, if the first-stage features only include features x1, x2, x3, and x4, then the first target prediction model trained by the first device training data corresponding to the features only including x1, x2, x3, and x4 is selected to perform a first order rate prediction on the target user of the first stage to obtain a first prediction intermediate result; or if the first-stage features only include features x1, x2, x3, x4, and x5, then the first target prediction model trained by the first device training data corresponding to the features only including x1, x2, x3, x4, and x5 is selected to perform a first order rate prediction on the target user of the first stage to obtain a first prediction intermediate result; or if the first-stage features only include features x1, x2, x3, x4, and x6, then the first target prediction model trained by the first device training data corresponding to the features only including x1, x2, x3, x4, and x6 is selected to perform a first order rate prediction on the target user of the first stage to obtain a first prediction intermediate result. The first target prediction model is used to predict the first order rate of the target users in the first stage to obtain a first prediction intermediate result, or if the first stage features only include the features of x1, x2, x3, x4, x5 and x7, then the first target prediction model trained with the first device training data corresponding to the features only including x1, x2, x3, x4, x5 and x7 is selected to predict the first order rate of the target users in the first stage to obtain a first prediction intermediate result, or if the first stage features only include the features of x1, x2, x3, x4, x5, x6 and x7, then the first target prediction model trained with the first device training data corresponding to the features only including x1, x2, x3, x4, x5, x6 and x7 is selected to predict the first order rate of the target users in the first stage to obtain a first prediction intermediate result.

[0090] Specifically, assume that the first-stage feature representation vector is (a, b), where a represents the user's age of 35, and b represents the user's hobby of traveling, etc. Further, the first-stage features are input into the first target prediction model, the first-stage feature representation vector is classified to obtain a classification result vector, and the classification result vector is used as the first prediction intermediate result.

[0091] It should be noted that, since the first target prediction model is already trained, it can accurately predict the first order rate of the target user and obtain the first prediction intermediate result. After obtaining the first prediction intermediate result, a vertical federation prediction is performed with the second device based on the first prediction intermediate result to obtain the aggregation result of the first stage.

[0092] The step of obtaining the first target prediction model of the first stage features includes:

[0093] Step A1, obtaining first device training data with preset labels of the first stage features, and obtaining second device training data of the second device in the first stage;

[0094] Step A2, obtaining a preset basic model, and performing federated iterative training on the preset basic model based on the first device training data with preset labels and the second device training data until a preset termination training condition is reached to obtain the first target prediction model.

[0095] In this embodiment, it should be noted that the first device and the second device need to perform vertical federated learning modeling to construct a vertical federated model jointly held by the first device and the second device, wherein the vertical federated model includes a preset first part prediction model (which can be a first target prediction model, and also includes other target prediction models, etc.) and a preset second part vertical federated model, wherein the preset first part prediction model is the part of the vertical federated model held by the first device, and the preset second part prediction model is the part of the vertical federated model held by the second device, and the number of the second devices is greater than or equal to 1, that is, the vertical federated learning modeling can be multi-party federated learning modeling or two-party federated learning modeling, wherein, if The first part of the prediction model is preset as the first target prediction model, and the method of constructing the vertical federated model jointly held by the first device and the second device includes: obtaining the first device training data with preset labels of the first stage characteristics, such as the first device training data corresponding to the features only including x1, x2, x3 and x4, and obtaining the second device training data of the second device in the first stage (including only the second device training data corresponding to the features a1, a2, a3 and a4), obtaining the preset basic model, and performing federal iterative training on the preset basic model based on the first device training data with preset labels and the second device training data until the preset termination training conditions are met, such as the convergence of the preset loss function, to obtain the first target prediction model.

[0096] The step of performing vertical federated prediction with the second device based on the first prediction intermediate result to obtain the first stage aggregation result includes:

[0097] Step B1, receiving a second prediction intermediate result of the second device behavior data of the target user in the first stage determined by the second device;

[0098] Step B2, aggregating the first prediction intermediate result and the second prediction intermediate result to obtain an aggregated result of the first stage.

[0099] After obtaining the first prediction intermediate result of the first-stage behavior data of the target user, receive the second prediction intermediate result of the second device behavior data of the target user in the first stage determined by the second device, aggregate the first prediction intermediate result and the second prediction intermediate result to obtain the aggregated result of the first stage. For example, assuming that the vertical federation model is a linear model Y=M1X1+M2X2+M3X3+M4X4+N1a1+N2a2+N3a3+N4a4, the first target prediction model is Y1=M1X1+M2X2+M3X3+M4X4, and the preset second part prediction model is Y2=N1a1+N2a2+N3a3+N4a4, and then the first prediction intermediate result is Y1, the second prediction intermediate result is Y2, and the aggregated result of the first stage is Y=(Y1+Y2) / 2. For example, assuming that the first prediction intermediate result is 0.8 and the second prediction intermediate result is 0.9, the first prediction intermediate result and the second prediction intermediate result are weighted averaged to obtain the aggregated result of the first stage of 0.85.

[0100] Step S20, based on the aggregation result of the first stage, determining the first stage interaction qualification level of the target user to determine the second stage behavior data;

[0101] Determine the first-stage interaction qualification level of the first-stage aggregation results to determine the second-stage behavior data, for example, determine the threshold interval corresponding to the first-stage aggregation results, and determine the first-stage interaction qualification level of the target user based on the threshold interval. For example, the first-stage aggregation result, that is, the order rate, is 0.3. At this time, the threshold interval corresponding to the first-stage aggregation result is the first threshold interval, and the corresponding first feedback information is common questions and answers (pre-stored). After obtaining the first feedback information, obtain the second-stage behavior data of the target user determined based on the first feedback information. Specifically, the user generates a (click or read) behavior based on the first feedback information, then the generated (click or read) behavior and the first-stage behavior data constitute the second-stage behavior data.

[0102] Step S30, based on the second stage behavior data, perform vertical federation prediction with the second device to obtain the second stage aggregation result of the target user to obtain the second stage interaction qualification level of the target user, wherein the second stage is a time-series continuous stage with the first stage.

[0103] In this embodiment, based on the second-stage behavior data, a vertical federation prediction is performed with the second device to obtain the second-stage aggregation result of the target user, so as to obtain the second-stage interaction qualification level of the target user, wherein the second stage is a stage that is sequentially continuous with the first stage, wherein the second stage is a stage that is sequentially continuous with the first stage, since the second stage is a stage that is sequentially continuous with the first stage, that is, based on the order rate in the previous stage (the aggregation result in the first stage), the interaction qualification level in the first stage is determined, and the first feedback information (service content) is further determined, and the order rate in the subsequent stage (the aggregation result in the second stage) is determined based on the first feedback information. That is, in this embodiment, the interaction qualification level of the first stage can be determined according to the prediction result (the order rate in the previous stage), and then the service content can be graded, and the service content or the first feedback data (determined by the interaction qualification level in the first stage) corresponding to the level can be provided. The user generates behavior in the service content or the first feedback data (determined by the interaction qualification level in the first stage), and then the prediction of the next stage is carried out, and the richness of the sample data is expanded based on the vertical federation in each stage, that is, the present application obtains the target prediction model of the corresponding stage in stages or grades to predict user behavior, and the richness of the sample data is expanded based on the vertical federation in each stage, thereby improving the prediction accuracy of the order rate.

[0104] In this embodiment, specifically, a second target prediction model of the second stage characteristics is obtained (obtained by training with the first device training data including only x1, x2, x3, x4 and x5 features (second stage characteristics)), and based on the second target prediction model, a second order rate prediction is performed for the target users of the second stage to obtain the aggregation results of the second stage.

[0105] Based on the aggregation results of the second stage, a second stage interaction qualification level is obtained, based on the second stage interaction qualification level, a second interaction file is determined, based on the second interaction file, the interaction mode of the second stage (such as machine interaction mode or manual interaction mode) and second feedback data (second feedback information or service content) are determined, so that the target user can perform behavioral operations in the next stage based on the interaction mode and second feedback data of the second stage, such as performing behavioral operations in the third stage based on the interaction mode and second feedback data of the second stage.

[0106] The present application obtains the first prediction intermediate result of the first-stage behavior data of the target user, and performs vertical federation prediction with the second device based on the first prediction intermediate result to obtain the first-stage aggregation result; based on the first-stage aggregation result, the first-stage interaction qualification level of the target user is determined to determine the second-stage behavior data; based on the second-stage behavior data, the vertical federation prediction is performed with the second device to obtain the second-stage aggregation result of the target user to obtain the second-stage interaction qualification level of the target user, wherein the second stage and the first stage are time-series continuous stages. Compared with the prior art that determines the user's interaction qualification level by determining the order rate prediction result of the behavior data distributed in a specific dimension, the present application can achieve the user order rate prediction result of the corresponding stage by stage or grade, and then obtain the target user's interaction qualification level, and in the present application, by vertical federation with other devices, the richness of the data is broadened, so that the order rate is predicted based on richer behavior data, thereby improving the prediction accuracy of the order rate prediction result, and then improving the determination accuracy of the target user's interaction qualification level, overcoming the technical defect of low accuracy of determination of the target user's interaction qualification level in the prior art.

[0107] Further, based on the first embodiment of the present application, another embodiment of the present application is provided. In this embodiment, referring to Figure 2 The step of determining the first-stage interaction qualification level of the first-stage interaction qualification level determination result comprises:

[0108] Step S21, determining the first-stage interaction qualification level of the target user based on the aggregation result of the first stage;

[0109] Step S22, determining a first interaction file based on the first-stage interaction qualification level;

[0110] Step S23, parsing the first interaction file to determine the second stage behavior data.

[0111] Based on the numerical size of the aggregation result of the first stage (order rate), such as 0.3, the corresponding threshold interval is determined to determine the first-stage interaction qualification level of the target user in the first stage. Specifically, if the order rate (aggregation result of the first stage) is divided into three threshold intervals in the interaction qualification level determination system, where the first threshold interval is (0, 0.33), the second threshold interval is (0.33-0.66), and the third threshold interval is (0.66-1), then the first-stage interaction qualification level of the aggregation result (order rate) of the first stage is the first level.

[0112] Determine the first interaction file of the first stage interaction qualification level, parse the first interaction file, obtain the first interaction file including the interaction method and the content of the first feedback information, and then determine the target user to obtain the second stage behavior data based on the content of the first feedback information.

[0113] Specifically, for example, the interaction file may be: a preset common marketing article, a pure machine interaction mode, and output or feedback common questions and answers after the interaction;

[0114] Or the interaction file can be: preset machine-selected articles, machine-first then human interaction, output or feedback of popular questions and answers after interaction;

[0115] Or the interaction file can be: preset manually selected articles, pure manual interaction mode, output or feedback of manually selected questions and answers after interaction, etc.

[0116] After obtaining the first interaction file (first feedback data), the user generates further behavior based on the first interaction file (first feedback data), and obtains second-stage behavior data based on the further behavior and the first-stage behavior data.

[0117] The step of parsing the first interaction file to determine the second-stage behavior data includes:

[0118] Step C1, determining the channel type of the target user;

[0119] Step C2, parsing the first interaction file, and determining second-stage behavior data based on the channel type and the parsed first interaction file.

[0120] Determine the first interaction file of the first-stage interaction qualification level. After determining the first interaction file, determine the channel type of the target user, such as through WeChat or APP, or through phone calls, text messages, etc. Based on the channel type and the first feedback data determined by the first interaction file, determine the second-stage behavior data.

[0121] In this embodiment, the first feedback data and the second-stage behavior data are determined by the channel type and the first interaction file, thereby avoiding resource consumption caused by attempting to output the first feedback data through multiple channels.

[0122] In this embodiment, the first-stage interaction qualification level of the target user is determined based on the aggregation result of the first stage; the first interaction file is determined based on the first-stage interaction qualification level; and the first interaction file is parsed to determine the second-stage behavior data. In this embodiment, the second-stage behavior number is accurately determined based on the first-stage interaction qualification level.

[0123] In another embodiment of the method for determining the interaction qualification level of the present application, after the step of performing vertical federation prediction with the second device based on the second-stage behavior data to obtain the second-stage aggregated result of the target user, and using the aggregated result as the second-stage order rate, the method includes:

[0124] Step S40: If it is detected that the order rate prediction results of the target user in multiple time-series continuous stages are all greater than a preset value, a preset special identification process is performed on the target user;

[0125] Step S50, executing a preset targeted recommendation process for the target user after the preset special identification processing.

[0126] In this embodiment, if it is detected that the order rate prediction results of the target user in multiple time-series continuous stages are all greater than the preset value, it indicates that the target user is a user who needs to be recommended, and thus a preset special identification processing is performed on the target user.

[0127] A preset targeted recommendation process is executed for the target users who have been processed with the preset special identification. The preset targeted recommendation process may include: fixed-point push to the target users who have been processed with the preset special identification, or telephone push to the target users who have been processed with the preset special identification. To avoid causing user disgust, if no user feedback is received after a preset number of fixed-point pushes or telephone pushes, no push will be made within a preset time period.

[0128] In this embodiment, if it is detected that the order rate prediction results of the target user in multiple time-series continuous stages are all greater than the preset value, the target user is processed with a preset special mark; and a preset targeted recommendation process is executed for the target user after the preset special mark processing. In this embodiment, targeted services are provided to users with preset special mark processing to improve user experience.

[0129] Further, based on the above embodiment of the present application, another embodiment of the present application is provided. In this embodiment, the step of determining the first-stage interaction qualification level of the target user based on the aggregation result of the first stage to determine the second-stage behavior data includes:

[0130] Step a1, determining the first-stage interaction qualification level of the first-stage target user based on the aggregation result of the first stage;

[0131] Step a2, obtaining the channel type of the first-stage interaction qualification level;

[0132] Step a3: if the channel type is a public account channel, outputting preset selected articles to the target user based on the first level and the public account channel;

[0133] Step a4, obtaining the second-stage behavior data of the target user determined based on the preset selected articles.

[0134] Specifically, for example, the aggregation result of the first stage, that is, the order rate, is 0.3. At this time, the channel type of the target user is obtained. If the channel type is a public account channel, the corresponding first feedback information is to output a preset selected article to the target user. After obtaining the preset selected article output to the target user, the second stage behavior data of the target user determined based on the preset selected article is obtained. Specifically, the user generates a (click to read, read for half an hour) behavior based on the preset selected article. The generated (click to read, read for half an hour) behavior and the first stage behavior data constitute the second stage behavior data, and then the second feedback information is determined based on the second stage behavior data. In this embodiment, the second stage behavior data is accurately determined.

[0135] Reference Figure 3 , Figure 3 It is a schematic diagram of the device structure of the hardware operating environment involved in the embodiment of the present application.

[0136] like Figure 3 As shown, the interactive qualification level determination device may include: a processor 1001, such as a CPU, a memory 1005, and a communication bus 1002. The communication bus 1002 is used to realize the connection and communication between the processor 1001 and the memory 1005. The memory 1005 may be a high-speed RAM memory or a stable memory (non-volatile memory), such as a disk memory. The memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0137] Optionally, the interactive qualification level determination device may also include a rectangular user interface, a network interface, a camera, an RF (Radio Frequency) circuit, a sensor, an audio circuit, a WiFi module, etc. The rectangular user interface may include a display screen (Display), an input submodule such as a keyboard (Keyboard), and the optional rectangular user interface may also include a standard wired interface and a wireless interface. The network interface may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface).

[0138] Those skilled in the art will understand that Figure 3 The structure of the interactive qualification level determination device shown in the figure does not constitute a limitation on the interactive qualification level determination device, and may include more or less components than those shown in the figure, or combine certain components, or arrange the components differently.

[0139] like Figure 3As shown, the memory 1005 as a storage medium may include an operating system, a network communication module and an interactive qualification level determination program. The operating system is a program that manages and controls the hardware and software resources of the interactive qualification level determination device, and supports the operation of the interactive qualification level determination program and other software and / or programs. The network communication module is used to realize the communication between the components inside the memory 1005, and the communication with other hardware and software in the interactive qualification level determination system.

[0140] exist Figure 3 In the interactive qualification level determination device shown, the processor 1001 is used to execute the interactive qualification level determination program stored in the memory 1005 to implement the steps of any of the interactive qualification level determination methods described above.

[0141] The specific implementation of the interactive qualification level determination device of the present application is basically the same as the embodiments of the interactive qualification level determination method described above, and will not be repeated here.

[0142] The present application also provides an interactive qualification level determination device, the interactive qualification level determination device comprising:

[0143] A first acquisition module, used to acquire a first prediction intermediate result of the first-stage behavior data of the target user, and perform a vertical federated prediction with the second device based on the first prediction intermediate result to obtain an aggregated result of the first stage;

[0144] A determination module, configured to determine the first-stage interaction qualification level of the target user based on the aggregation result of the first stage, so as to determine the second-stage behavior data;

[0145] The second acquisition module is used to perform longitudinal federated prediction with the second device based on the second-stage behavior data to obtain the second-stage aggregation result of the target user to obtain the second-stage interaction qualification level of the target user, wherein the second stage is a time-series continuous stage with the first stage.

[0146] Optionally, the first acquisition module includes:

[0147] A first acquisition unit, configured to acquire first-stage behavior data of a target user and determine first-stage features of the first-stage behavior data;

[0148] The second acquisition unit is used to acquire a first target prediction model of the first stage characteristics, and based on the first target prediction model, perform a first order rate prediction on the target user to obtain a first prediction intermediate result.

[0149] Optionally, the second acquiring unit includes:

[0150] A first acquisition subunit is used to acquire first device training data with preset labels of the first stage features, and acquire second device training data of the second device in the first stage;

[0151] The second acquisition subunit is used to acquire a preset basic model, and based on the first device training data with preset labels and the second device training data, perform federated iterative training on the preset basic model until a preset termination training condition is reached to obtain the first target prediction model.

[0152] Optionally, the first acquisition module includes:

[0153] A receiving unit, configured to receive a second prediction intermediate result of the second device behavior data of the target user in the first stage determined by the second device;

[0154] An aggregation unit is used to aggregate the first prediction intermediate result and the second prediction intermediate result to obtain an aggregation result of the first stage.

[0155] Optionally, the determining module includes:

[0156] A third acquisition unit, configured to acquire the aggregation result based on the first stage, and determine the first stage interaction qualification level of the target user;

[0157] A first determining unit, configured to determine a first interaction file based on the first-stage interaction qualification level;

[0158] The second determining unit is used to parse the first interaction file and determine the second stage behavior data.

[0159] Optionally, the second determining unit includes:

[0160] A first determining subunit, configured to determine a channel type of the target user;

[0161] The second determining subunit is configured to parse the first interaction file and determine the second-stage behavior data based on the channel type and the parsed first interaction file.

[0162] Optionally, the interaction qualification level determination device further includes:

[0163] A detection module, configured to perform a preset special identification process on the target user if it is detected that the order rate prediction results of the target user in multiple time-series continuous stages are all greater than a preset value;

[0164] The execution module is used to execute a preset targeted recommendation process for the target user after the preset special identification processing.

[0165] Optionally, the determining module includes:

[0166] A third determining unit, configured to determine a first-stage interaction qualification level of a first-stage target user based on the aggregation result of the first stage;

[0167] A fourth obtaining unit, configured to obtain the channel type of the first-stage interaction qualification level;

[0168] A fourth determining unit, configured to determine, if the channel type is a public account channel, to output a preset selected article to the target user based on the first level and the public account channel;

[0169] The fifth acquisition unit is used to acquire the second-stage behavior data of the target user determined based on the preset selected articles.

[0170] The specific implementation of the interactive qualification level determination device of the present application is basically the same as the various embodiments of the interactive qualification level determination method described above, and will not be repeated here.

[0171] An embodiment of the present application provides a storage medium, and the storage medium stores one or more programs, and the one or more programs can also be executed by one or more processors to implement the steps of any of the above-mentioned methods for determining the interaction qualification level.

[0172] The specific implementation of the storage medium of the present application is basically the same as the above-mentioned embodiments of the interaction qualification level determination method, and will not be repeated here.

[0173] It should be noted that, in this article, the terms "include", "comprises" or any other variations 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 sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0174] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0175] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0176] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for determining an interaction qualification level, characterized in that: The interactive qualification level determination method comprises: Obtaining a first prediction intermediate result of the first stage behavior data of the target user, and performing a vertical federated prediction with the second device based on the first prediction intermediate result to obtain an aggregated result of the first stage; Based on the aggregation result of the first stage, determining the first stage interaction qualification level of the target user to determine the second stage behavior data, the second stage behavior data includes the first stage behavior data and the behavior data generated by the target user based on feedback information, the feedback information is the service content fed back by the enterprise to the target user based on the first stage interaction qualification level, and the behavior data generated based on the feedback information is the operation behavior data generated by the user when interacting with the service content recorded by the enterprise; Based on the second-stage behavior data, a vertical federated prediction is performed with the second device to obtain the second-stage aggregation result of the target user to obtain the second-stage interaction qualification level of the target user, wherein the second stage and the first stage are time-series continuous stages, and the aggregation results of the first stage and the aggregation results of the second stage are both the order rate of the target user.

2. The interactive qualification level determination method according to claim 1, characterized in that: The step of obtaining the first prediction intermediate result of the first stage behavior data of the target user includes: Acquire the first-stage behavior data of the target user, and determine the first-stage characteristics of the first-stage behavior data; A first target prediction model of the first stage features is obtained, and based on the first target prediction model, a first order rate prediction is performed on the target user to obtain a first prediction intermediate result.

3. The interactive qualification level determination method according to claim 2, characterized in that: The step of obtaining the first target prediction model of the first stage features includes: Acquire first device training data with preset labels for the first stage features, and acquire second device training data for the second device in the first stage; A preset basic model is obtained, and based on the first device training data with preset labels and the second device training data, the preset basic model is federatedly trained until a preset termination training condition is reached to obtain the first target prediction model.

4. The interactive qualification level determination method according to claim 1, characterized in that: The step of performing vertical federated prediction with the second device based on the first prediction intermediate result to obtain the first stage aggregation result includes: Receiving a second prediction intermediate result of the second device behavior data of the target user in the first stage determined by the second device; The first prediction intermediate result and the second prediction intermediate result are aggregated to obtain an aggregated result of the first stage.

5. The interactive qualification level determination method according to claim 1, characterized in that: The step of determining the first-stage interaction qualification level of the target user based on the aggregation result of the first stage to determine the second-stage behavior data includes: Determining the first-stage interaction qualification level of the target user based on the aggregation result of the first stage; determining a first interaction file based on the first-stage interaction qualification level; The first interaction file is parsed to determine the second-stage behavior data.

6. The interactive qualification level determination method according to claim 5, characterized in that: The step of parsing the first interaction file to determine the second-stage behavior data includes: Determine the channel type of the target user; The first interaction file is parsed, and second-stage behavior data is determined based on the channel type and the parsed first interaction file.

7. The method for determining the interactive qualification level according to any one of claims 1 to 6, characterized in that: After the step of performing longitudinal federated prediction with the second device based on the second-stage behavior data to obtain the second-stage aggregated result of the target user to obtain the second-stage interaction qualification level of the target user, the method includes: If it is detected that the order rate prediction results of the target user in multiple time-series continuous stages are all greater than the preset value, a preset special identification processing is performed on the target user; A preset targeted recommendation process is executed for the target user after the preset special identification processing.

8. The interactive qualification level determination method according to claim 1, characterized in that: The step of determining the first-stage interaction qualification level of the target user based on the aggregation result of the first stage to determine the second-stage behavior data includes: Determining the first-stage interaction qualification level of the first-stage target user based on the aggregation result of the first stage; Obtain the channel type of the first-stage interaction qualification level; If the channel type is a public account channel, outputting a preset selected article to the target user based on the first level and the public account channel; The second-stage behavior data of the target user determined based on the preset selected articles is obtained.

9. An interactive qualification level determination device, characterized in that: The interactive qualification level determination device comprises: A first acquisition module, used to acquire a first prediction intermediate result of the first-stage behavior data of the target user, and perform a vertical federated prediction with the second device based on the first prediction intermediate result to obtain an aggregated result of the first stage; a determination module, for determining the first-stage interaction qualification level of the target user based on the aggregation result of the first stage, so as to determine the second-stage behavior data, wherein the second-stage behavior data includes the first-stage behavior data and the behavior data generated by the target user based on feedback information, wherein the feedback information is the service content fed back by the enterprise to the target user based on the first-stage interaction qualification level, and the behavior data generated based on the feedback information is the operation behavior data generated by the user when interacting with the service content and recorded by the enterprise; The second acquisition module is used to perform vertical federated prediction with the second device based on the second-stage behavior data to obtain the second-stage aggregation result of the target user to obtain the second-stage interaction qualification level of the target user, wherein the second stage and the first stage are time-series continuous stages, and the aggregation results of the first stage and the aggregation results of the second stage are both the order rate of the target user.

10. An interactive qualification level determination device, characterized in that: The interactive qualification level determination device comprises: a memory, a processor, and a program stored in the memory for implementing the interactive qualification level determination method. The memory is used to store a program for implementing the interactive qualification level determination method; The processor is used to execute a program for implementing the interaction qualification level determination method, so as to implement the steps of the interaction qualification level determination method according to any one of claims 1 to 8.

11. A storage medium, characterized in that: The storage medium stores a program for implementing the interactive qualification level determination method, and the program for implementing the interactive qualification level determination method is executed by a processor to implement the steps of the interactive qualification level determination method according to any one of claims 1 to 8.

Images