Personalized subscription page generation method and device, electronic equipment and storage medium

Abstract

The invention provides a personalized subscription page generation method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring multi-dimensional behavior data generated when a user uses a target application program; extracting behavior characteristics corresponding to the user from the multi-dimensional behavior data; constructing a behavior portrait corresponding to the user according to the behavior characteristics; predicting a renewing willingness level corresponding to the user according to the behavior characteristics; and determining a corresponding renewal welfare allocation strategy according to the behavior portrait and the renewal willingness level, and generating a personalized subscription page according to the renewal welfare allocation strategy. According to the method and the device, the renewal welfare allocation strategy suitable for the user can be determined in combination with the user behavior portrait and the renewal willingness, and the personalized subscription page is generated according to the renewal welfare allocation strategy to stimulate the user to renew, so that the retention rate of the user is improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method, device, electronic device and storage medium for generating a personalized subscription page. Background Art

[0002] With the rapid popularization of video streaming and other online content consumption services, major platforms have launched membership subscription services to attract and retain users. The membership renewal issue has gradually become one of the key points for optimizing platform operations and improving user stickiness. In the prior art, membership renewal reminders are generally sent to users by pushing a fixed subscription page.

[0003] However, this generalized push strategy uses the same solution for different users, lacks personalization, is difficult to motivate users to renew their subscriptions, and has a high user churn rate. Summary of the invention

[0004] The purpose of the embodiments of the present application is to provide a personalized subscription page generation method, device, electronic device and storage medium to solve the problem that the generalized push strategy adopts the same solution for different users, lacks personalization, and easily disturbs users, and the current member renewal reminder lacks effective user benefit distribution and is difficult to motivate users to renew. The specific technical solution is as follows:

[0005] In a first aspect, the present application provides a method for generating a personalized subscription page, comprising:

[0006] Obtain multi-dimensional behavioral data generated by users when using the target application;

[0007] Extracting the behavior characteristics corresponding to the user from the multi-dimensional behavior data;

[0008] Constructing a behavior profile corresponding to the user according to the behavior characteristics;

[0009] Predicting the renewal willingness level of the user according to the behavior characteristics;

[0010] A corresponding renewal benefit allocation strategy is determined according to the behavior portrait and the renewal willingness level, and a personalized subscription page is generated according to the renewal benefit allocation strategy.

[0011] In a possible implementation, predicting the renewal willingness level of the user according to the behavior characteristics includes:

[0012] The behavior feature is input into a pre-trained renewal willingness prediction model, so that the renewal willingness prediction model outputs the renewal willingness level corresponding to the user.

[0013] In a possible implementation, the method further includes:

[0014] Get sample data;

[0015] Using the sample data to train several machine learning algorithms to obtain an initial prediction model corresponding to each of the machine algorithms;

[0016] For each initial prediction model, cross-validate the initial prediction model to obtain a first validation result, and validate the initial prediction model using a validation set to obtain a second validation result;

[0017] Determining a model score corresponding to the initial prediction model according to the first verification result and the second verification result;

[0018] The initial prediction model with the highest corresponding model score is determined as the renewal intention prediction model.

[0019] In a possible implementation, the method further includes:

[0020] Obtaining actual feedback data corresponding to the renewal willingness prediction model;

[0021] Adding the actual feedback data to the sample data to obtain new sample data;

[0022] The renewal intention prediction model is trained using the new sample data to optimize the renewal intention prediction model.

[0023] In a possible implementation, before cross-validating the initial prediction model to obtain a first validation result and validating the initial prediction model using a validation set to obtain a second validation result, the method further includes:

[0024] Obtaining a set of model parameter combinations corresponding to the initial prediction model, wherein the set of model parameter combinations includes all parameter combinations corresponding to the initial prediction model obtained through grid search, or a partial parameter combination corresponding to the initial prediction model obtained through random search;

[0025] For each model parameter combination, cross-validate the model parameter combination to obtain a parameter validation score;

[0026] The model parameter combination with the highest corresponding parameter verification score is used as the target model parameter combination for the initial prediction model application.

[0027] In a possible implementation, extracting the behavior features corresponding to the user from the multi-dimensional behavior data includes:

[0028] Preprocessing the multi-dimensional behavior data;

[0029] Convert the preprocessed multi-dimensional behavior data into periodic behavior data;

[0030] The behavior features corresponding to the user are extracted from the periodic behavior data.

[0031] In a possible implementation, extracting the behavior feature corresponding to the user from the periodic behavior data includes:

[0032] extracting a number of initial features from the periodic behavior data;

[0033] determining a feature importance score for each of the initial features;

[0034] Sorting all the initial features in descending order of the corresponding feature importance scores;

[0035] A preset number of initial features before sorting are selected as the behavioral features corresponding to the user.

[0036] In a second aspect, the present application provides a personalized subscription page generation device, comprising:

[0037] An acquisition module is used to acquire multi-dimensional behavior data generated by users when using the target application;

[0038] An extraction module, used to extract the behavior features corresponding to the user from the multi-dimensional behavior data;

[0039] A construction module, used to construct a behavior profile corresponding to the user according to the behavior characteristics;

[0040] A prediction module, used to predict the renewal willingness level of the user according to the behavior characteristics;

[0041] An allocation module is used to determine a corresponding renewal benefit allocation strategy according to the behavior portrait and the renewal willingness level, and generate a personalized subscription page according to the renewal benefit allocation strategy.

[0042] In a possible implementation, the prediction module is specifically used to:

[0043] The behavior feature is input into a pre-trained renewal willingness prediction model, so that the renewal willingness prediction model outputs the renewal willingness level corresponding to the user.

[0044] In a possible implementation, the device further includes a training module, which is used to:

[0045] Get sample data;

[0046] Using the sample data to train several machine learning algorithms to obtain an initial prediction model corresponding to each of the machine algorithms;

[0047] For each initial prediction model, cross-validate the initial prediction model to obtain a first validation result, and validate the initial prediction model using a validation set to obtain a second validation result;

[0048] Determining a model score corresponding to the initial prediction model according to the first verification result and the second verification result;

[0049] The initial prediction model with the highest corresponding model score is determined as the renewal intention prediction model.

[0050] In a possible implementation, the training module is further used to:

[0051] Obtaining actual feedback data corresponding to the renewal willingness prediction model;

[0052] Adding the actual feedback data to the sample data to obtain new sample data;

[0053] The renewal intention prediction model is trained using the new sample data to optimize the renewal intention prediction model.

[0054] In a possible implementation, the training module is further used to:

[0055] Obtaining a set of model parameter combinations corresponding to the initial prediction model, wherein the set of model parameter combinations includes all parameter combinations corresponding to the initial prediction model obtained through grid search, or a partial parameter combination corresponding to the initial prediction model obtained through random search;

[0056] For each model parameter combination, cross-validate the model parameter combination to obtain a parameter validation score;

[0057] The model parameter combination with the highest corresponding parameter verification score is used as the target model parameter combination for the initial prediction model application.

[0058] In a possible implementation, the extraction module is specifically used to:

[0059] Preprocessing the multi-dimensional behavior data;

[0060] Convert the preprocessed multi-dimensional behavior data into periodic behavior data;

[0061] The behavior features corresponding to the user are extracted from the periodic behavior data.

[0062] In a possible implementation, the extraction module is further used to:

[0063] extracting a number of initial features from the periodic behavior data;

[0064] determining a feature importance score for each of the initial features;

[0065] Sorting all the initial features in descending order of the corresponding feature importance scores;

[0066] A preset number of initial features before sorting are selected as the behavioral features corresponding to the user.

[0067] In a third aspect, an electronic device is provided, comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus;

[0068] Memory, used to store computer programs;

[0069] The processor is used to implement any method step described in the first aspect when executing a program stored in the memory.

[0070] In a fourth aspect, a computer-readable storage medium is provided, characterized in that a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, any method step described in the first aspect is implemented.

[0071] In a fifth aspect, a computer program product comprising instructions is provided, which, when executed on a computer, enables the computer to execute any of the above-mentioned methods for generating a personalized subscription page.

[0072] Beneficial effects of the embodiments of the present application:

[0073] The embodiments of the present application provide a method, device, electronic device and storage medium for generating a personalized subscription page. The present application first obtains the multi-dimensional behavior data generated by the user when using the target application, and extracts the behavior characteristics corresponding to the user from the multi-dimensional behavior data. Then, a behavior portrait corresponding to the user is constructed based on the behavior characteristics, and the renewal willingness level corresponding to the user is predicted based on the behavior characteristics. Finally, a corresponding renewal benefit allocation strategy is determined based on the behavior portrait and the renewal willingness level, and a personalized subscription page is generated based on the renewal benefit allocation strategy. Through the present application, a renewal benefit allocation strategy suitable for the user can be determined in combination with the user's behavior portrait and renewal willingness, and a personalized subscription page is generated based on the renewal benefit allocation strategy to motivate the user to renew, thereby improving the user's retention rate.

[0074] Of course, implementing any product or method of the present application does not necessarily require achieving all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0076] In order to more clearly illustrate the embodiments of the present invention 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.

[0077] One or more embodiments are exemplarily described by pictures in the corresponding drawings, and these exemplified descriptions do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0078] Figure 1 A flowchart of a method for generating a personalized subscription page provided in an embodiment of the present application;

[0079] Figure 2 A flow chart of data collection provided in an embodiment of the present application;

[0080] Figure 3 A flow chart of an embodiment of model training provided in an embodiment of the present application;

[0081] Figure 4 A processing flow chart of a method for generating a personalized subscription page provided in an embodiment of the present application;

[0082] Figure 5 A schematic diagram of the structure of a personalized subscription page generation device provided in an embodiment of the present application;

[0083] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0084] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0085] The disclosure below provides many different embodiments or examples to implement different structures of the present invention. In order to simplify the disclosure of the present invention, the parts and settings of specific examples are described below. Of course, they are only examples, and the purpose is not to limit the present invention. In addition, the present invention can repeat reference numbers and / or letters in different examples. This repetition is for the purpose of simplification and clarity, and does not itself indicate the relationship between the various embodiments and / or settings discussed.

[0086] Figure 1 A flow chart of a personalized subscription page generation method provided for an embodiment of the present application. This method can be applied to one or more electronic devices such as smart phones, laptops, desktop computers, portable computers, servers, etc. In addition, the execution subject of this method can be hardware or software. When the above-mentioned execution subject is hardware, the execution subject can be one or more of the above-mentioned electronic devices. For example, a single electronic device can execute this method, or multiple electronic devices can cooperate with each other to execute this method. When the above-mentioned execution subject is software, this method can be implemented as multiple software or software modules, or as a single software or software module. It is not specifically limited here.

[0087] like Figure 1 As shown, the method specifically includes:

[0088] Step 101: Obtain multi-dimensional behavior data generated by the user when using the target application.

[0089] Target applications refer to applications that require renewal reminders, such as video playback applications, shopping applications, etc.

[0090] Multi-dimensional behavior data refers to the behavior data of multiple dimensions generated by users when using the target application, such as the user's historical viewing time, viewing frequency, comments and likes, and other interactive behavior data, as well as search keywords, device type used, viewing time period, last renewal time, etc. In the application, multi-dimensional behavior data can be collected through different clients (such as PC, mobile, smart TV, etc.).

[0091] In actual applications, the multi-dimensional behavior data generated by each user when using the target application can be collected and stored in real time, and the time when the behavior data is generated can be recorded, thereby obtaining the corresponding time series data.

[0092] Based on this, in the embodiment of the present application, the time series data of each user in the current time period can be obtained from the pre-collected data at regular intervals as the current multi-dimensional behavior data, and the current user behavior profile analysis and renewal willingness prediction can be performed. Thus, the corresponding renewal benefits can be issued according to the user's real-time behavior data.

[0093] For ease of understanding, the real-time data collection process (taking a video playback application as an example) is described below. Figure 2 As shown, the following steps are included:

[0094] (1) Pingback delivery:

[0095] When a user watches a video, likes, comments, collects, shares, etc. while using a video playback application, the client will immediately send a Pingback signal to the server to record the current user's specific operations. Pingback signals generally contain data such as user ID, video ID, operation type, operation time, and device information.

[0096] (2) RocketMQ collection:

[0097] Pingback signals are first delivered to a message queue system (such as RocketMQ) to ensure the real-time and reliability of data. RocketMQ achieves high-throughput data transmission and supports sequential storage and consumption of messages, making it suitable for the collection of large-scale user behavior data.

[0098] (3) Spark real-time processing:

[0099] After the data enters the message queue, you can use a real-time computing framework (such as Apache Spark) to process and pre-analyze the data. Spark Streaming accesses RocketMQ's data stream to process and analyze user behavior data in real time, and clean and aggregate the data. Data preprocessing includes steps such as data format standardization, deduplication, missing value filling, and noise data filtering.

[0100] (4)Hive offline storage:

[0101] The data processed in real time will be stored in a distributed file system (such as HDFS) and stored in a data warehouse (such as Hive) in a table format. Hive allows the use of the SQL-like HQL language to query and analyze data, facilitating subsequent batch data processing and modeling.

[0102] Step 102: extracting the behavior features corresponding to the user from the multi-dimensional behavior data.

[0103] In the embodiment of the present application, step 102 may specifically include the following steps:

[0104] Step A1, preprocessing the multi-dimensional behavior data;

[0105] Step A2, converting the pre-processed multi-dimensional behavior data into periodic behavior data;

[0106] Step A3: extracting the behavior features corresponding to the user from the periodic behavior data.

[0107] In the application, the initial multi-dimensional behavior data includes behavior data at a certain moment, for example, at a certain time on a certain day of a certain year, user A collected video B, at a certain time on a certain day of a certain year, user C liked video D, or at a certain time on a certain day of a certain year, user E watched video using device F, and so on.

[0108] Periodic behavior data refers to behavior data within a period (such as daily, weekly, monthly, etc.) based on multi-dimensional behavior data statistics. For example, user A watches the program twice a day, user B watches the program 10 times a month, and so on.

[0109] In this implementation, first, the collected raw data is preprocessed, including cleaning, filtering, and standardizing the data, including removing noise data, processing missing values ​​and outliers, etc. Then, the preprocessed data is converted into periodic behavior data, and the user's behavior characteristics are extracted from the periodic behavior data.

[0110] Compared with directly collected time series data, periodic behavior data can remove some noise data, and has the advantages of simplifying analysis difficulty, discovering long-term advantages, and improving prediction accuracy.

[0111] Specifically, extracting the behavioral features corresponding to the user from the periodic behavioral data may include the following steps: extracting a number of initial features from the periodic behavioral data, determining a feature importance score for each of the initial features, sorting all of the initial features in descending order of the corresponding feature importance scores, and selecting a preset number of initial features before sorting as the behavioral features corresponding to the user.

[0112] Feature importance score, which is used to measure the contribution of each initial feature to the prediction result.

[0113] In the application, the feature importance score of each initial feature can be determined as follows:

[0114] 1. Decision Trees and Tree-Based Models:

[0115] (1) Decision Trees: measure feature importance by the error (such as impurity or information gain) reduced when splitting a node. (2) Random Forests and XGBoost: accumulate feature importance scores across all trees.

[0116] 2. Linear Model:

[0117] (1) Linear regression and logistic regression: Evaluate by the absolute value of the feature coefficient (weight). The larger the coefficient, the more important the feature. Alternatively, using regularization (such as Lasso regression) will make the coefficients of unimportant features zero, thereby screening out important features.

[0118] 3. LIME (Local Interpretable Model-Agnostic Explanations): Evaluate the importance of features through local explanation models.

[0119] 4. SHAP (SHapley Additive exPlanations); calculate the Shapley value of each feature to quantify its contribution to the prediction results.

[0120] 5. Feature selection techniques: (1) Recursive feature elimination (RFE): repeatedly train the model and remove the least important features. (2) Statistical measurement: use methods such as chi-square test and mutual information to evaluate the importance of features.

[0121] Through this solution, the most important features can be obtained, the model can be simplified, and the performance can be improved.

[0122] Step 103: construct a behavior profile corresponding to the user according to the behavior characteristics.

[0123] Behavior profile is a comprehensive description of user behavior, which can include user interest tags, usage preferences and behavior patterns. Among them, interest tags: for example, the types of movies that users prefer to watch (action movies, comedy movies, etc.), and their favorite stars. Device preferences: on which devices users prefer to watch movies (mobile phones, computers, tablets, etc.). Behavior patterns: for example, at what time (morning, noon, evening) do users tend to watch movies, and the viewing frequency per week / month, etc.

[0124] For example, the behavior profile of user A may include information: he watches comedy movies on his mobile phone at around 8 o'clock every evening, about 30 times a month; the behavior profile of user B may include information: she watches action movies on her tablet every weekend, about 10 times a month.

[0125] Step 104: predicting the user's corresponding renewal willingness level according to the behavior characteristics.

[0126] The renewal willingness level is used to characterize the user's renewal possibility. The higher the level, the higher the user's renewal possibility. In the application, the renewal willingness level can include: highly likely to renew, medium likely to renew, low likely to renew, and very likely to churn users.

[0127] In the embodiment of the present application, step 104 may specifically include the following steps:

[0128] The behavior feature is input into a pre-trained renewal willingness prediction model, so that the renewal willingness prediction model outputs the renewal willingness level corresponding to the user.

[0129] In this way, the renewal intention prediction model can be used to intelligently predict the user's renewal intention level based on behavioral characteristics.

[0130] Step 105: Determine a corresponding renewal benefit allocation strategy based on the behavior profile and the renewal willingness level, and generate a personalized subscription page based on the renewal benefit allocation strategy.

[0131] In an embodiment of the present application, a corresponding renewal benefit allocation strategy is determined by combining the user's behavioral profile and renewal willingness level, and a personalized subscription page is generated according to the renewal benefit allocation strategy to distribute corresponding renewal benefits to the user.

[0132] For example, for users who may churn, their historical consumption records and discount usage are evaluated based on their behavioral profiles, and suitable renewal order coupons are recommended to them through a personalized subscription page. For users who are highly likely to renew and whose behavioral profiles show high-frequency viewing, they are provided with renewal coupons with smaller discounts through a personalized subscription page to incentivize them to continue subscribing, and so on.

[0133] In an embodiment of the present application, first, the multi-dimensional behavior data generated by the user when using the target application is obtained, and the behavior characteristics corresponding to the user are extracted from the multi-dimensional behavior data. Then, the behavior portrait corresponding to the user is constructed based on the behavior characteristics, and the renewal willingness level corresponding to the user is predicted based on the behavior characteristics. Finally, the corresponding renewal benefit allocation strategy is determined based on the behavior portrait and the renewal willingness level, and a personalized subscription page is generated based on the renewal benefit allocation strategy. Through this application, a renewal benefit allocation strategy suitable for the user can be determined in combination with the user's behavior portrait and renewal willingness, and a personalized subscription page is generated based on the renewal benefit allocation strategy to encourage users to renew, thereby improving the user's retention rate.

[0134] See also Figure 3 , is a flow chart of an embodiment of model training provided in the embodiment of the present application. Figure 3 As shown, the process may include the following steps:

[0135] Step 301: Obtain sample data.

[0136] Sample data refers to historical behavior features extracted after preprocessing the historical behavior data of users on the target application, and these historical behavior features are labeled according to whether the user renews or churns.

[0137] Step 302: Use the sample data to train several machine learning algorithms to obtain an initial prediction model corresponding to each of the machine algorithms.

[0138] The above-mentioned machine learning algorithms may include random forest algorithm, XGBoost algorithm, LSTM (Long Short-Term Memory), GBDT (Gradient Boosting Decision Tree), SVM (Support Vector Machine) algorithm, Logistic Regression algorithm, Neural Networks algorithm, etc.

[0139] In the embodiment of the present application, first, the sample data is divided into data sets, divided into training sets, validation sets, and test sets, and then these data sets are used to train various machine learning algorithms to obtain several initial prediction models.

[0140] Step 303: for each initial prediction model, cross-validate the initial prediction model to obtain a first validation result, and validate the initial prediction model using a validation set to obtain a second validation result.

[0141] In one embodiment, the cross-validation process is: divide the data set into multiple parts, and train and validate each part, so as to ensure that the model can perform well on different data sets. Among them, K-fold cross-validation is the most commonly used method. Taking 5-fold cross-validation (K=5) to evaluate the model as an example, it specifically includes the following steps:

[0142] (1) All member data are randomly divided into five subsets of equal size, denoted as D1, D2, D3, D4, and D5.

[0143] (2) Perform K training and validation. In the first iteration, use D1 as the validation set, D2, D3, D4, and D5 as the training set, train the model, and evaluate the performance on D1. In the second iteration, use D2 as the validation set, D1, D3, D4, and D5 as the training set, train the model, and evaluate the performance on D2. Repeat this process until all subsets have been used as validation sets once.

[0144] (3) Calculate the performance index for each iteration

[0145] In each iteration, the confusion matrix, ROC curve, AUC and other performance indicators are calculated. For example: Accuracy; Precision; Recall; F1 Score; AUC (Area Under the Curve); Then, the performance indicators of all K iterations are averaged to obtain the final model evaluation results, for example:

[0146] Average accuracy = (Acc1+Acc2+Acc3+Acc4+Acc5) / 5

[0147] Average precision = (Prec1+Prec2+Prec3+Prec4+Prec5) / 5

[0148] Average recall rate = (Recall1+Recall2+Recall3+Recall4+Recall5) / 5

[0149] Average F1 score = (F1_1+F1_2+F1_3+F1_4+F1_5) / 5

[0150] Average AUC = (AUC1 + AUC2 + AUC3 + AUC4 + AUC5) / 5

[0151] By calculating the average performance index of the cross-validation results through the above scheme, that is, the first validation result, it is possible to determine which model performs stably and excellently on all folds, thereby selecting the model with the best performance for actual deployment.

[0152] In one embodiment, the validation set is a portion of data extracted from the original data, which is not used in the model training process and is used to test the model performance.

[0153] As a specific example, the process of using the validation set for validation is as follows: data of 2,000 members are randomly selected from the total data of 10,000 members as the validation set, and multiple trained machine learning models (such as random forest, logistic regression, and gradient boosting decision tree) are predicted on the validation set, that is, the validation set is input into each trained model to obtain the prediction result of each model.

[0154] Then, the model performance indicators are calculated based on the prediction results. The specific process is as follows:

[0155] (1) Confusion Matrix:

[0156] Definition: Displays the number of correct and incorrect predictions made by the model, constructed by calculating the following four values:

[0157] True Positives (TP): The number of users that the model correctly predicts will renew their subscription.

[0158] True Negatives (TN): The number of users that the model correctly predicts will not renew their subscription.

[0159] False Positives (FP): The number of users that the model incorrectly predicts will renew their subscription (in fact, they will not renew their subscription).

[0160] False Negatives (FN): The number of users that the model incorrectly predicts will not renew their subscription (in fact, they will renew their subscription).

[0161] Confusion matrix example (taking the random forest model as an example):

[0162] The model predicts that 500 users will renew their subscriptions, but 400 users actually do renew their subscriptions (TP=400), and 100 users actually do not renew their subscriptions (FP=100).

[0163] The model predicted that 1500 users would not renew their subscriptions, but 1300 users did not renew their subscriptions (TN=1300), and 200 users actually renewed their subscriptions (FN=200).

[0164] (2) Calculate the following performance indicators based on the confusion matrix:

[0165] Accuracy: The proportion of correct predictions made by the model.

[0166] Calculation formula: (TP+TN) / (TP+TN+FP+FN).

[0167] Actual calculation example: (400+1300) / (400+1300+100+200)=1700 / 2000=0.85.

[0168] Precision: The percentage of customers correctly predicted to renew their subscription.

[0169] Calculation formula: TP / (TP+FP).

[0170] Actual calculation example: 400 / (400+100)=400 / 500=0.80.

[0171] Recall rate: The proportion of users who are correctly predicted to renew their subscriptions.

[0172] Calculation formula: TP / (TP+FN)

[0173] Actual calculation example: 400 / (400+200)=400 / 600=0.67

[0174] F1 score: The harmonic mean of precision and recall.

[0175] Calculation formula: 2*(Precision*Recall) / (Precision+Recall)

[0176] Actual calculation example: 2*(0.80*0.67) / (0.80+0.67)≈0.73

[0177] (3) ROC curve and AUC value:

[0178] ROC curve (Receiver Operating Characteristic Curve): used to depict the false positive rate and true positive rate of the model under different thresholds, showing the classification ability of the model.

[0179] AUC (Area Under the Curve): refers to the area under the ROC curve. An AUC value close to 1 indicates superior model performance. The AUC of the random forest model is 0.85.

[0180] Then, a second verification result is obtained by weighted average calculation.

[0181] Step 304: Determine a model score corresponding to the initial prediction model according to the first verification result and the second verification result.

[0182] Step 305: Determine the initial prediction model with the highest corresponding model score as the renewal intention prediction model.

[0183] For ease of understanding, steps 304 and 305 are described in a unified manner as follows:

[0184] In the embodiment of the present application, after obtaining the first verification result and the second verification result, a model score is obtained by performing a weighted average calculation on the first verification result and the second verification result, and the initial prediction model with the highest corresponding model score is determined as the renewal intention prediction model. In this way, it can be ensured that the renewal intention prediction model performs stably and excellently in all folds, and has excellent comprehensive performance in all indicators.

[0185] In addition, in another embodiment of the present application, the method may further include the following steps:

[0186] Acquire actual feedback data corresponding to the renewal intention prediction model, add the actual feedback data to the sample data to obtain new sample data, and use the new sample data to train the renewal intention prediction model to optimize the renewal intention prediction model.

[0187] The above-mentioned actual feedback data refers to actual user behavior data. In the scenario of member renewal and churn prediction, the feedback data includes but is not limited to the user's actual renewal behavior.

[0188] In an embodiment of the present application, after the renewal intention prediction model is put into use, user feedback data is collected in real time, and the model is updated and iteratively optimized through the feedback data to improve the long-term effect of the model.

[0189] In addition, in another embodiment of the present application, the method may further include the following steps: obtaining a set of model parameter combinations corresponding to the initial prediction model, wherein the set of model parameter combinations includes all parameter combinations corresponding to the initial prediction model obtained by grid search, or, partial parameter combinations corresponding to the initial prediction model obtained by random search, and for each model parameter combination, cross-validating the model parameter combination to obtain a parameter validation score, and using the model parameter combination with the highest corresponding parameter validation score as the target model parameter combination for the application of the initial prediction model.

[0190] In the embodiments of the present application, a grid search or a random search combined with a cross-validation method is used to debug the model parameters to improve the prediction effect of the model.

[0191] Optionally, the embodiment of the present application also provides a processing flow for allocating renewal benefits, such as Figure 4 As shown, the specific steps are as follows.

[0192] (1) User behavior data: user viewing history, interactive behavior, device information, etc. on the target application. This data is collected and stored by the system.

[0193] (2) Data processing:

[0194] Data preprocessing: Clean, filter, and standardize the collected raw data, including removing noise data, processing missing values ​​and outliers, etc.

[0195] Feature engineering: Extract multi-dimensional features from preprocessed data, including viewing time, frequency, number of interactions, device usage trends, time characteristics, interest tags, etc., and convert them into periodic behavioral characteristics to generate feature vectors for model training.

[0196] (3) Model training and evaluation:

[0197] Model training: Use machine learning algorithms such as random forest to train feature data and generate a prediction model.

[0198] Hyperparameter tuning: Use grid search or random search to adjust model parameters and optimize model performance.

[0199] Cross-validation: Use the K-fold cross-validation method to evaluate the performance of the model on different data sets to prevent overfitting.

[0200] Model validation: Evaluate model performance on the validation set, use confusion matrix, ROC curve and other indicators to select the best model.

[0201] (4) Model deployment:

[0202] Model deployment: The best model that has been trained and verified is deployed to the online system to predict users' renewal willingness in real time.

[0203] Model prediction: Use the deployed model to predict users' renewal willingness based on their real-time data, and provide support for subsequent personalized push and benefit distribution.

[0204] Feedback loop: Collect user feedback data, continuously monitor model performance, and update and iteratively optimize the model through feedback data to improve the long-term effectiveness of the model.

[0205] Through this application, it is possible to combine the user's behavioral profile and renewal intention to determine a renewal benefit allocation strategy suitable for the user, and generate a personalized subscription page based on the renewal benefit allocation strategy to encourage users to renew, thereby improving user retention rate.

[0206] Based on the same technical concept, the embodiment of the present application also provides a personalized subscription page generation device, such as Figure 5 As shown, the device comprises:

[0207] An acquisition module 51 is used to acquire multi-dimensional behavior data generated by the user when using the target application;

[0208] An extraction module 52, configured to extract the behavior features corresponding to the user from the multi-dimensional behavior data;

[0209] A construction module 53 is used to construct a behavior profile corresponding to the user according to the behavior characteristics;

[0210] Prediction module 54, used for predicting the renewal willingness level corresponding to the user according to the behavior characteristics;

[0211] The allocation module 55 is used to determine the corresponding renewal benefit allocation strategy according to the behavior profile and the renewal willingness level, and generate a personalized subscription page according to the renewal benefit allocation strategy.

[0212] In a possible implementation, the prediction module is specifically used to:

[0213] The behavior feature is input into a pre-trained renewal willingness prediction model, so that the renewal willingness prediction model outputs the renewal willingness level corresponding to the user.

[0214] In a possible implementation, the device further includes a training module, which is used to:

[0215] Get sample data;

[0216] Using the sample data to train several machine learning algorithms to obtain an initial prediction model corresponding to each of the machine algorithms;

[0217] For each initial prediction model, cross-validate the initial prediction model to obtain a first validation result, and validate the initial prediction model using a validation set to obtain a second validation result;

[0218] Determining a model score corresponding to the initial prediction model according to the first verification result and the second verification result;

[0219] The initial prediction model with the highest corresponding model score is determined as the renewal intention prediction model.

[0220] In a possible implementation, the training module is further used to:

[0221] Obtaining actual feedback data corresponding to the renewal willingness prediction model;

[0222] Adding the actual feedback data to the sample data to obtain new sample data;

[0223] The renewal intention prediction model is trained using the new sample data to optimize the renewal intention prediction model.

[0224] In a possible implementation, the training module is further used to:

[0225] Obtaining a set of model parameter combinations corresponding to the initial prediction model, wherein the set of model parameter combinations includes all parameter combinations corresponding to the initial prediction model obtained through grid search, or a partial parameter combination corresponding to the initial prediction model obtained through random search;

[0226] For each model parameter combination, cross-validate the model parameter combination to obtain a parameter validation score;

[0227] The model parameter combination with the highest corresponding parameter verification score is used as the target model parameter combination for the initial prediction model application.

[0228] In a possible implementation, the extraction module is specifically used to:

[0229] Preprocessing the multi-dimensional behavior data;

[0230] Convert the preprocessed multi-dimensional behavior data into periodic behavior data;

[0231] The behavior features corresponding to the user are extracted from the periodic behavior data.

[0232] In a possible implementation, the extraction module is further used to:

[0233] extracting a number of initial features from the periodic behavior data;

[0234] determining a feature importance score for each of the initial features;

[0235] Sorting all the initial features in descending order of the corresponding feature importance scores;

[0236] A preset number of initial features before sorting are selected as the behavioral features corresponding to the user.

[0237] In an embodiment of the present application, first, the multi-dimensional behavior data generated by the user when using the target application is obtained, and the behavior characteristics corresponding to the user are extracted from the multi-dimensional behavior data. Then, the behavior portrait corresponding to the user is constructed based on the behavior characteristics, and the renewal willingness level corresponding to the user is predicted based on the behavior characteristics. Finally, the corresponding renewal benefit allocation strategy is determined based on the behavior portrait and the renewal willingness level, and a personalized subscription page is generated based on the renewal benefit allocation strategy. Through this application, a renewal benefit allocation strategy suitable for the user can be determined in combination with the user's behavior portrait and renewal willingness, and a personalized subscription page is generated based on the renewal benefit allocation strategy to encourage users to renew, thereby improving the user's retention rate.

[0238] Based on the same technical concept, the embodiment of the present application also provides an electronic device, such as Figure 6 As shown, it includes a processor 111, a communication interface 112, a memory 113 and a communication bus 114, wherein the processor 111, the communication interface 112, and the memory 113 communicate with each other through the communication bus 114.

[0239] Memory 113, used for storing computer programs;

[0240] The processor 111 is used to execute the program stored in the memory 113 to implement the following steps:

[0241] Obtain multi-dimensional behavioral data generated by users when using the target application;

[0242] Extracting the behavior characteristics corresponding to the user from the multi-dimensional behavior data;

[0243] Constructing a behavior profile corresponding to the user according to the behavior characteristics;

[0244] Predicting the renewal willingness level of the user according to the behavior characteristics;

[0245] A corresponding renewal benefit allocation strategy is determined according to the behavior portrait and the renewal willingness level, and a personalized subscription page is generated according to the renewal benefit allocation strategy.

[0246] The communication bus mentioned in the above electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0247] The communication interface is used for communication between the above electronic device and other devices.

[0248] The memory may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0249] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0250] In another embodiment provided in the present application, a computer-readable storage medium is provided, in which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned personalized subscription page generation methods are implemented.

[0251] In another embodiment provided by the present application, a computer program product including instructions is also provided, which, when executed on a computer, enables the computer to execute any of the personalized subscription page generation methods in the above embodiments.

[0252] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0253] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a general hardware platform, and of course, by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the relevant technology can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0254] It should be understood that the terms used herein are only for the purpose of describing specific example embodiments and are not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms "one", "an" and "said" as used herein may also be meant to include plural forms. The terms "include", "comprise", "contain", and "have" are inclusive, and therefore specify the existence of stated features, steps, operations, elements and / or parts, but do not exclude the existence or addition of one or more other features, steps, operations, elements, parts, and / or combinations thereof. The method steps, processes, and operations described herein are not interpreted as necessarily requiring them to be performed in the specific order described or illustrated, unless the execution order is clearly indicated. It should also be understood that additional or alternative steps may be used.

[0255] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for generating a personalized subscription page, characterized in that: The method comprises: Obtain multi-dimensional behavioral data generated by users when using the target application; Extracting the behavior characteristics corresponding to the user from the multi-dimensional behavior data; Constructing a behavior profile corresponding to the user according to the behavior characteristics; Predicting the renewal willingness level of the user according to the behavior characteristics; A corresponding renewal benefit allocation strategy is determined according to the behavior portrait and the renewal willingness level, and a personalized subscription page is generated according to the renewal benefit allocation strategy.

2. The method according to claim 1, characterized in that The predicting the renewal willingness level of the user according to the behavior characteristics includes: The behavior feature is input into a pre-trained renewal willingness prediction model, so that the renewal willingness prediction model outputs the renewal willingness level corresponding to the user.

3. The method according to claim 2, characterized in that The method further comprises: Get sample data; Using the sample data to train several machine learning algorithms to obtain an initial prediction model corresponding to each of the machine algorithms; For each initial prediction model, cross-validate the initial prediction model to obtain a first validation result, and validate the initial prediction model using a validation set to obtain a second validation result; Determining a model score corresponding to the initial prediction model according to the first verification result and the second verification result; The initial prediction model with the highest corresponding model score is determined as the renewal intention prediction model.

4. The method according to claim 3, characterized in that The method further comprises: Obtaining actual feedback data corresponding to the renewal willingness prediction model; Adding the actual feedback data to the sample data to obtain new sample data; The renewal intention prediction model is trained using the new sample data to optimize the renewal intention prediction model.

5. The method according to claim 3, characterized in that: Before cross-validating the initial prediction model to obtain a first validation result and validating the initial prediction model using a validation set to obtain a second validation result, the method further includes: Obtaining a set of model parameter combinations corresponding to the initial prediction model, wherein the set of model parameter combinations includes all parameter combinations corresponding to the initial prediction model obtained through grid search, or a partial parameter combination corresponding to the initial prediction model obtained through random search; For each model parameter combination, cross-validate the model parameter combination to obtain a parameter validation score; The model parameter combination with the highest corresponding parameter verification score is used as the target model parameter combination for the initial prediction model application.

6. The method according to claim 1, characterized in that The extracting the behavior features corresponding to the user from the multi-dimensional behavior data includes: Preprocessing the multi-dimensional behavior data; Convert the preprocessed multi-dimensional behavior data into periodic behavior data; The behavior features corresponding to the user are extracted from the periodic behavior data.

7. The method according to claim 6, characterized in that The extracting the behavior feature corresponding to the user from the periodic behavior data includes: extracting a number of initial features from the periodic behavior data; determining a feature importance score for each of the initial features; Sorting all the initial features in descending order of the corresponding feature importance scores; A preset number of initial features before sorting are selected as the behavioral features corresponding to the user.

8. A personalized subscription page generation device, characterized in that: The device comprises: An acquisition module is used to acquire multi-dimensional behavior data generated by users when using the target application; An extraction module, used to extract the behavior features corresponding to the user from the multi-dimensional behavior data; A construction module, used to construct a behavior profile corresponding to the user according to the behavior characteristics; A prediction module, used to predict the renewal willingness level of the user according to the behavior characteristics; An allocation module is used to determine a corresponding renewal benefit allocation strategy according to the behavior portrait and the renewal willingness level, and generate a personalized subscription page according to the renewal benefit allocation strategy.

9. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; Memory, used to store computer programs; The processor is used to implement the personalized subscription page generation method described in any one of claims 1-7 when executing the program stored in the memory.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for generating a personalized subscription page according to any one of claims 1 to 7 is implemented.

Images