User portrait generation method and system based on user behaviors

Abstract

The invention discloses a user portrait generation method and system based on user behaviors, and relates to the technical field of communication, and the method comprises the steps: collecting package attention data of a target user on a specific website and application; obtaining a to-be-predicted package attention feature based on the package attention data; collecting package use data of the target user; obtaining a to-be-predicted package use feature based on the package use data; feedback evaluation data of the target user are collected, and the feedback evaluation data comprise scoring information, evaluation information and feedback period information; obtaining feedback evaluation features to be predicted based on the feedback evaluation data; the to-be-predicted package attention features, the to-be-predicted package use features and the to-be-predicted feedback evaluation features are input into the user portrait generation model to obtain a user portrait corresponding to the target user, and the user portrait comprises a target recommended package type and target package recommendation time.

Description

Technical Field

[0001] This specification relates to the field of communication technology, and more specifically, the present application relates to a method and system for generating a user portrait based on user behavior. Background Art

[0002] With the rapid development of information technology, especially the progress in 5G communications, big data and artificial intelligence, consumers' expectations for smartphones are also increasing. Therefore, choosing the right mobile phone package has become particularly critical. Most of the mobile phone packages on the market are designed based on the experience of operators, which leads to many packages being unscientific and not meeting the needs of rational consumers, thus affecting the user experience and brand loyalty.

[0003] In order to improve the design of mobile phone packages and enhance user experience and loyalty, operators can create detailed user portraits based on user behavior data. Therefore, it is necessary to propose an intelligent and accurate user portrait generation method and system based on user behavior. Summary of the invention

[0004] A series of simplified concepts are introduced in the Summary of the Invention section, which will be further described in detail in the Detailed Description of the Invention section. The Summary of the Invention section of this application does not mean to attempt to define the key features and essential technical features of the claimed technical solution, nor does it mean to attempt to determine the scope of protection of the claimed technical solution.

[0005] In a first aspect, the present application proposes a method for generating a user profile based on user behavior, the method comprising:

[0006] Collect package attention data of target users on specific websites and applications, wherein the package attention data includes package type information, package browsing frequency information, package browsing cumulative duration information and package purchase history information;

[0007] Based on the above package attention data, the package attention features to be predicted are obtained;

[0008] Collecting the package usage data of the target users, wherein the package usage data includes data usage, call usage and communication records of frequent contacts;

[0009] Obtaining usage characteristics of the package to be predicted based on the above package usage data;

[0010] Collecting feedback and evaluation data of the target users, wherein the feedback and evaluation data includes rating information, evaluation information and feedback cycle information;

[0011] Acquire feedback evaluation features to be predicted based on the above feedback evaluation data;

[0012] The above-mentioned package attention features to be predicted, the above-mentioned package usage features to be predicted and the above-mentioned feedback evaluation features to be predicted are input into the above-mentioned user portrait generation model to obtain the user portrait corresponding to the above-mentioned target user, wherein the above-mentioned user portrait includes the target recommended package type and the target package recommendation time.

[0013] In a feasible implementation manner, the step of obtaining the package attention features to be predicted based on the package attention data includes:

[0014] Calculate the package preference characteristics of each package type by weighted summation method according to the package browsing frequency information and the package purchase history information;

[0015] According to the above package browsing cumulative time information, the package attention characteristics of each package type are extracted by time-weighted average method;

[0016] According to the above package browsing frequency information and the above package browsing cumulative duration information, the K-means algorithm is used to analyze the distribution characteristics of user package browsing time.

[0017] In a feasible implementation manner, the obtaining the usage characteristics of the package to be predicted based on the package usage data includes:

[0018] Calculate the usage level characteristics based on the above data usage and the above call usage;

[0019] The usage characteristics during break time and the usage characteristics during working hours are extracted based on the communication records of the above-mentioned frequent contacts.

[0020] In a feasible implementation manner, the step of obtaining the feedback evaluation features to be predicted based on the feedback evaluation data includes:

[0021] Determine the package satisfaction characteristics through statistical analysis based on the above rating information;

[0022] Determine the emotional attitude characteristics based on the above evaluation information through natural language processing method;

[0023] Based on the above feedback cycle information, the time series analysis method is used to determine the feedback timeliness characteristics.

[0024] In a feasible implementation manner, the specific steps of the method for constructing the above user portrait generation model include:

[0025] A first training subset is constructed based on historical package attention features, a second training subset is constructed based on historical package usage features, a first training subset is constructed based on historical feedback evaluation features, and a fourth training subset is constructed based on user portraits corresponding to historical target users;

[0026] Constructing a first decision tree corresponding to the package attention feature, a second decision tree corresponding to the package usage feature, and a third decision tree corresponding to the feedback evaluation feature;

[0027] Training the first decision tree using the first training subset;

[0028] Training the second decision tree using the second training subset;

[0029] Training the third decision tree using the third training subset;

[0030] Additional decision trees are added incrementally based on the gradient boosting method to correct the prediction residuals;

[0031] The fourth training subset, the first decision tree, the second decision tree, the third decision tree, the additional decision tree, the voting module and the weighted average module are integrated to form the user portrait generation model, wherein the voting module is used to determine the target recommended package type by voting according to the output information of all decision trees, and the weighted average module is used to determine the target package recommendation time by weighted average method according to the output information of all decision trees.

[0032] In a feasible implementation manner, the specific steps of the above-mentioned method for optimizing the user portrait generation model include:

[0033] The depth and learning rate of the decision tree are determined as parameters to be optimized;

[0034] Calculate the first target parameter combination under the first parameter optimization mode and the second target parameter combination under the second parameter optimization mode for all parameter combinations to be optimized;

[0035] An optimal target parameter combination is determined according to the first target parameter combination or the second target parameter combination.

[0036] In a feasible implementation manner, the method for determining the first target parameter combination in the first parameter optimization mode includes:

[0037] Perform model training and validation operations on each set of parameter combinations to be optimized to obtain performance characteristics of the validation data set, wherein the performance characteristics include accuracy and AUC;

[0038] A first target parameter combination is selected based on the above performance characteristics to complete a first parameter optimization mode.

[0039] In a feasible implementation manner, the method for determining the second target parameter combination in the second parameter optimization mode includes:

[0040] Use Gaussian process to build a probability model between the parameter combination to be optimized and the model performance;

[0041] Calculate the performance loss of the above-mentioned parameter combination to be optimized in the current group according to the above-mentioned probability model;

[0042] The probability model is updated based on the performance loss and a new set of parameter combinations to be optimized, and the performance loss is calculated again based on the updated probability model;

[0043] Repeat the probability model update operation and the performance loss calculation operation to a preset number of times or until the performance parameter improvement is less than a preset threshold, and determine the corresponding target parameter combination at this time as the above-mentioned second target parameter combination to complete the second parameter optimization mode.

[0044] In a feasible implementation manner, determining the optimal target parameter combination according to the first target parameter combination and the second target parameter combination includes:

[0045] Get the package promotion target;

[0046] When the package promotion target is the fast popular target, the first target parameter combination is determined as the optimal target parameter combination;

[0047] When the package promotion target is the maximum market target, the second target parameter combination is used to determine the optimal target parameter combination.

[0048] In the second aspect, the present application proposes a user profile generation system based on user behavior, including:

[0049] A first collecting unit is used to collect package attention data of target users on specific websites and applications, wherein the package attention data includes package type information, package browsing frequency information, package browsing cumulative duration information and package purchase history information;

[0050] A first acquisition unit, configured to acquire the package attention feature to be predicted based on the package attention data;

[0051] A second collecting unit is used to collect the package usage data of the target user, wherein the package usage data includes data usage, call usage and communication records of frequent contacts;

[0052] A second acquisition unit, configured to acquire the usage characteristics of the package to be predicted based on the package usage data;

[0053] A third collecting unit is used to collect feedback evaluation data of the target user, wherein the feedback evaluation data includes scoring information, evaluation information and feedback cycle information;

[0054] A third acquisition unit, configured to acquire the feedback evaluation feature to be predicted based on the feedback evaluation data;

[0055] The fourth acquisition unit is used to input the above-mentioned package attention features to be predicted, the above-mentioned package usage features to be predicted and the above-mentioned feedback evaluation features to be predicted into the above-mentioned user portrait generation model to obtain the user portrait corresponding to the above-mentioned target user, wherein the above-mentioned user portrait includes the target recommended package type and the target package recommendation time.

[0056] In summary, the user profile generation method based on user behavior of the embodiment of the present application includes: collecting package attention data of the target user on specific websites and applications, wherein the package attention data includes package type information, package browsing frequency information, package browsing cumulative time information and package purchase history information; obtaining package attention features to be predicted based on the package attention data; collecting package usage data of the target user, wherein the package usage data includes data usage, call usage and communication records of frequent contacts; obtaining package usage features to be predicted based on the package usage data; collecting feedback evaluation data of the target user, wherein the feedback evaluation data includes rating information, evaluation information and feedback cycle information; obtaining feedback evaluation features to be predicted based on the feedback evaluation data; inputting the package attention features to be predicted, the package usage features to be predicted and the feedback evaluation features to be predicted into the user portrait generation model to obtain the user portrait corresponding to the target user, wherein the user portrait includes the target recommended package type and the target package recommended time. The method proposed in the present application generates a user portrait by analyzing the user's package attention data, package usage data and feedback evaluation data in detail, which can more accurately reflect the specific needs and preferences of each user. In this way, the recommended packages can better meet the personalized needs of users, thereby improving user satisfaction and increasing user stickiness. User portraits generated based on user behavior data can help operators allocate network and service resources more effectively. For example, for high-traffic users, operators can optimize relevant network configurations to ensure service quality, thereby improving the efficiency and quality of overall services. Through in-depth understanding of user behavior, marketing activities can be more targeted, thereby reducing ineffective marketing and advertising expenses. Operators can use user portraits to push the most suitable packages in a targeted manner, reduce promotion costs, and improve conversion rates. Accurate generation of user portraits and recommendation of personalized packages can significantly improve user brand loyalty. Users who feel that the service is personalized and focused on their needs are more likely to stay with the same operator for a long time. By continuously collecting user feedback and evaluation data and feeding this data into the user portrait generation model, continuous optimization and updating of services can be achieved, and the timeliness and relevance of services can be maintained. In summary, the user portrait generation method based on user behavior provided by this solution can not only meet user needs more accurately, but also significantly improve the service quality and economic benefits of operators. This method ensures the rationality and scientificity of package design through scientific data analysis, and overcomes the shortcomings of existing technologies.

[0057] The user behavior-based user portrait generation method proposed in this application and other advantages, objectives and features of this application will be partially reflected through the following description, and will also be partially understood by technical personnel in this field through research and practice of this application. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present specification. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0059] Figure 1 A schematic diagram of a user profile generation method based on user behavior provided in an embodiment of the present application;

[0060] Figure 2 A schematic diagram of the structure of a user portrait generation system based on user behavior provided in an embodiment of the present application. DETAILED DESCRIPTION

[0061] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. The technical solutions 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 only part of the embodiments of the present application, rather than all of the embodiments.

[0062] See also Figure 1 , Figure 1 A schematic diagram of a user profile generation method based on user behavior provided in an embodiment of the present application may specifically include:

[0063] S110, collecting package attention data of target users on specific websites and applications, wherein the package attention data includes package type information, package browsing frequency information, package browsing cumulative duration information and package purchase history information;

[0064] Exemplarily, package attention data of target users on specific websites and applications are collected.

[0065] Package focus data includes:

[0066] Package type information: records the different package types that users are interested in, such as unlimited data packages, family packages, or youth packages.

[0067] Package browsing frequency information: counts the number of times users visit each package type.

[0068] Cumulative package browsing time information: the cumulative time users spend on each package page.

[0069] Package purchase history information: records the packages that users have purchased in the past and their frequency.

[0070] S120, obtaining the package attention features to be predicted based on the package attention data;

[0071] Exemplarily, by analyzing package type information, package browsing frequency information, package browsing cumulative duration information, and package purchase history information, features describing user package preferences and behaviors are generated.

[0072] S130, collecting package usage data of the target user, wherein the package usage data includes data usage, call usage and communication records of frequent contacts;

[0073] Exemplarily, the package usage data of the target user is collected, and the package usage data includes:

[0074] Data usage: Records the user's data consumption, such as the amount of data traffic used per month.

[0075] Call usage: Statistics on users’ call duration and call frequency.

[0076] Frequently contacted communication records: records the communication details between the user and frequently contacted people, such as family, friends or colleagues.

[0077] S140, obtaining usage characteristics of the package to be predicted based on the package usage data;

[0078] Exemplarily, data usage, call usage and communication records of frequent contacts are analyzed to generate features that reflect the user's communication needs and preferences.

[0079] S150, collecting feedback evaluation data of the target user, wherein the feedback evaluation data includes rating information, evaluation information and feedback cycle information;

[0080] Exemplarily, feedback evaluation data of target users is collected, and the feedback evaluation data includes:

[0081] Rating information: The rating given by users to the packages they use.

[0082] Evaluation information: user’s textual evaluation of the package.

[0083] Feedback cycle information: records the time and frequency of user feedback

[0084] S160, obtaining feedback evaluation features to be predicted based on the feedback evaluation data;

[0085] Exemplarily, the rating information and evaluation information are analyzed, and natural language processing technology is used to extract the user's emotional attitude and satisfaction; at the same time, the feedback cycle information is evaluated to understand the user's activity and feedback habits.

[0086] S170. Input the above-mentioned package attention features to be predicted, the above-mentioned package usage features to be predicted and the above-mentioned feedback evaluation features to be predicted into the above-mentioned user portrait generation model to obtain the user portrait corresponding to the above-mentioned target user, wherein the above-mentioned user portrait includes the target recommended package type and the target package recommendation time.

[0087] Exemplarily, the above-mentioned package attention features to be predicted, package usage features to be predicted, and feedback evaluation features to be predicted are input into the user portrait generation model. The generated user portrait includes the target recommended package type and the target package recommendation time, based on the user's preferences, usage behavior, and feedback evaluation.

[0088] In summary, the method proposed in this application generates a user portrait that can more accurately reflect the specific needs and preferences of each user by analyzing the user's package attention data, package usage data, and feedback evaluation data in detail. In this way, the recommended package can better meet the personalized needs of users, thereby improving user satisfaction and increasing user stickiness. User portraits generated based on user behavior data can help operators allocate network and service resources more effectively. For example, for high-traffic users, operators can optimize relevant network configurations to ensure service quality, thereby improving the efficiency and quality of overall services. Through a deep understanding of user behavior, marketing activities can be more targeted, thereby reducing ineffective marketing and advertising expenses. Operators can use user portraits to push the most suitable packages in a targeted manner, reduce promotion costs, and increase conversion rates. The accurate generation of user portraits and the recommendation of personalized packages can significantly enhance user brand loyalty. Users feel that the service is personalized and focused on their needs, and are more likely to stay with the same operator for a long time. By continuously collecting user feedback evaluation data and feeding this data back to the user portrait generation model, continuous optimization and updating of services can be achieved, and the timeliness and relevance of services can be maintained. In summary, the user portrait generation method based on user behavior provided by this solution can not only meet user needs more accurately, but also significantly improve the service quality and economic benefits of operators. This method ensures the rationality and scientific nature of package design through scientific data analysis, overcoming the shortcomings of existing technologies.

[0089] In some examples, the step of obtaining the package attention features to be predicted based on the package attention data includes:

[0090] Calculate the package preference characteristics of each package type by weighted summation method according to the package browsing frequency information and the package purchase history information;

[0091] According to the above package browsing cumulative time information, the package attention characteristics of each package type are extracted by time-weighted average method;

[0092] According to the above package browsing frequency information and the above package browsing cumulative duration information, the K-means algorithm is used to analyze the distribution characteristics of user package browsing time.

[0093] Exemplarily, the package preference feature of each package type is calculated based on the package browsing frequency information and the package purchase history information.

[0094] Assign weights to browsing frequency and purchase history to reflect the importance of these indicators in preference judgment. Usually, purchase history is given a higher weight because it directly reflects purchase behavior. The calculation formula can be: Package preference feature = (browsing frequency × weight 1) + (purchase history × weight 2). This calculation provides a quantitative preference indicator that indicates the user's overall preference for a specific package type.

[0095] The package attention feature of each package type is extracted based on the cumulative package browsing time information. The browsing time is time-weighted averaged, and the recent browsing time may be given a higher weight to reflect the latest user interest. The calculation formula can be: Package attention feature = Σ(each browsing time × time weight) / total number of browsing times. This method can quantify the user's attention to a specific package, and a longer browsing time usually indicates a higher interest.

[0096] The K-means algorithm is used to analyze the time distribution characteristics of users browsing packages based on the package browsing frequency information and the cumulative package browsing time information. The time points (such as hours or minutes in a day) when users browse packages are collected to form a data set. The K-means clustering algorithm is used to group the time point data and identify the main browsing time periods. This analysis helps identify the peak time periods for users to browse packages and provide the recommendation system with a highly timely recommendation time.

[0097] Through the above methods, we can deeply understand the user's package attention behavior from different angles, and combine these features to build a more accurate user portrait, thereby optimizing the relevance and timeliness of package recommendations. The comprehensive use of these features provides a comprehensive view of user preferences and behaviors, providing data support for precision marketing and personalized services.

[0098] In some examples, the obtaining of the package usage features to be predicted based on the package usage data includes:

[0099] Calculate the usage level characteristics based on the above data usage and the above call usage;

[0100] The usage characteristics during break time and the usage characteristics during working hours are extracted based on the communication records of the above-mentioned frequent contacts.

[0101] For example, the user's data traffic usage and call duration are counted. Based on the user's total data usage and total call duration, the user is classified into different usage levels, such as "high usage", "medium usage" or "low usage". The usage level feature reflects the user's activity level in data and calls, and helps to identify different levels of user needs.

[0102] Analyze the communication records of frequent contacts to extract usage characteristics during rest periods and working hours. Monitor and record the time when users communicate with frequent contacts (such as family, friends, or colleagues). Identify the user's communication patterns during non-working hours such as evenings or weekends. Identify the user's communication patterns during normal working hours (such as 9:00AM to 5:00PM Monday to Friday). These characteristics help understand the user's living and working habits, so that package recommendations are more in line with their actual usage.

[0103] The method provided in the embodiment of the present application analyzes the user's usage level, so that operators can more accurately classify users and provide customized services and packages for different categories of users. Understanding the user's usage patterns during work hours and rest periods can help service providers provide services at the most appropriate time, such as providing discounts or promotions during user active periods. Based on the user's usage habits and active time periods, operators can manage network resources more effectively, such as increasing bandwidth allocation during high-demand periods. Accurate package recommendations and optimized service hours can increase user satisfaction and loyalty and reduce churn.

[0104] In some examples, obtaining the feedback evaluation features to be predicted based on the feedback evaluation data includes:

[0105] Determine the package satisfaction characteristics through statistical analysis based on the above rating information;

[0106] Determine the emotional attitude characteristics based on the above evaluation information through natural language processing method;

[0107] Based on the above feedback cycle information, the time series analysis method is used to determine the feedback timeliness characteristics.

[0108] For example, the above rating information is analyzed using statistical analysis to determine the package satisfaction characteristics. Collect and organize the user's rating data on the package. Calculate statistical indicators such as the mean, median, and variance of the rating. Convert the statistical results into package satisfaction characteristics to reflect the user's overall satisfaction with the current package. This feature helps to identify the user's satisfaction level with a specific package and guide future package improvements and promotion strategies.

[0109] Apply natural language processing technology to analyze the above evaluation information to determine the emotional attitude characteristics. Text preprocessing includes removing irrelevant characters, stop word processing, stemming, etc. Use sentiment analysis tools (such as Sentiment Analysis) to evaluate the emotional tendency (positive, negative or neutral) and intensity in the evaluation. Convert the sentiment analysis results into emotional attitude characteristics to describe the user's emotional response to the package.

[0110] This feature helps understand the user's emotional connection and the emotional color of feedback, which is crucial for customer relationship management and marketing activities.

[0111] Use time series analysis to analyze the feedback cycle information above to determine the feedback timeliness feature. Collect data about the time points of user feedback. Perform time series analysis on the collected time point data to identify the frequency and regularity of feedback. Convert the time series analysis results into feedback timeliness features, indicating the speed and frequency of user feedback. This feature helps to understand user engagement and the timeliness of service feedback, which is very useful for improving customer service and response mechanisms.

[0112] The method proposed in the embodiment of the present application can help companies adjust their products and services more accurately to meet user needs by understanding package satisfaction, user's emotional attitude, and the timeliness of feedback. Emotional attitude features can help marketing teams better understand market sentiment and adjust advertising and promotional activities to better resonate with user emotions. Analysis of feedback timeliness features can help customer service teams optimize their feedback mechanisms, solve user problems faster, and improve user experience. The comprehensive application of these features can provide companies with a data-driven decision-making basis and promote strategic decisions based on actual user data.

[0113] In some examples, the specific steps of the method for constructing the user portrait generation model include:

[0114] A first training subset is constructed based on historical package attention features, a second training subset is constructed based on historical package usage features, a first training subset is constructed based on historical feedback evaluation features, and a fourth training subset is constructed based on user portraits corresponding to historical target users;

[0115] Constructing a first decision tree corresponding to the package attention feature, a second decision tree corresponding to the package usage feature, and a third decision tree corresponding to the feedback evaluation feature;

[0116] Training the first decision tree using the first training subset;

[0117] Training the second decision tree using the second training subset;

[0118] Training the third decision tree using the third training subset;

[0119] Additional decision trees are added incrementally based on the gradient boosting method to correct the prediction residuals;

[0120] The fourth training subset, the first decision tree, the second decision tree, the third decision tree, the additional decision tree, the voting module and the weighted average module are integrated to form the user portrait generation model, wherein the voting module is used to determine the target recommended package type by voting according to the output information of all decision trees, and the weighted average module is used to determine the target package recommendation time by weighted average method according to the output information of all decision trees.

[0121] For example, a special training subset is created for each historical feature set, and a fourth training subset is constructed for the user glide corresponding to the historical target user, ensuring the focus and pertinence of the data and providing an accurate basis for model training. Each decision tree specializes in processing a feature set, which can accurately simulate and predict the user's behavior and preferences in the corresponding field. Ensure that each decision tree can learn key information from its corresponding data set and form an effective prediction model.

[0122] Additional decision trees are gradually added using the gradient boosting method to correct the prediction residuals of the previous decision trees. By continuously optimizing and enhancing the model, the prediction error is reduced and the overall prediction accuracy is improved.

[0123] The outputs of all decision trees are integrated using voting modules and weighted average modules. The voting module is used to determine the target recommended package type. The weighted average module is used to determine the target package recommendation time. The advantages of multiple models are integrated, and the prediction results of different models are used to improve the stability and accuracy of predictions through voting and weighted average methods.

[0124] The method proposed in the embodiment of the present application integrates different decision trees, and the model can analyze user data from multiple perspectives, integrate different prediction results, and enhance the overall prediction ability and accuracy. The gradient boosting method enables the model to gradually adapt to complex data features and improve its adaptability to new situations. Accurate package type recommendations and appropriate recommendation times can improve user satisfaction and enhance user experience. Through precise data analysis and model prediction, more scientific decision support is provided to help business growth and customer management.

[0125] In some examples, the specific steps of the above-mentioned method for optimizing the user portrait generation model include:

[0126] The depth and learning rate of the decision tree are determined as parameters to be optimized;

[0127] Calculate the first target parameter combination under the first parameter optimization mode and the second target parameter combination under the second parameter optimization mode for all parameter combinations to be optimized;

[0128] An optimal target parameter combination is determined according to the first target parameter combination or the second target parameter combination.

[0129] For example, the depth of the decision tree and the learning rate are selected as the parameters to be optimized. The depth of the decision tree affects the complexity of the model and the risk of overfitting; the learning rate affects the speed of parameter update and the stability of the model during model training.

[0130] The performance of each set of parameters can be evaluated in the first mode using a system traversal method such as grid search; the second mode may use more advanced optimization algorithms such as Bayesian optimization.

[0131] By fine-tuning the key parameters of the decision tree, the accuracy and efficiency of the model on specific tasks can be significantly improved, reducing the risk of overfitting and underfitting. The selection of optimal parameters helps improve the model's ability to predict unseen data, which is particularly important in practical applications facing a changing data environment. By determining the optimal learning rate and tree depth, model training can be made more efficient and unnecessary waste of computing resources can be reduced. Combining the results of the two optimization modes, while ensuring model performance, parameters can be flexibly adjusted according to business needs and resource constraints, supporting more accurate and reliable data-driven decisions.

[0132] In some examples, the method for determining the first target parameter combination in the first parameter optimization mode includes:

[0133] Perform model training and validation operations on each set of parameter combinations to be optimized to obtain performance characteristics of the validation data set, wherein the performance characteristics include accuracy and AUC;

[0134] A first target parameter combination is selected based on the above performance characteristics to complete a first parameter optimization mode.

[0135] Exemplarily, each set of parameter combinations to be optimized undergoes model training and subsequent validation. The performance of the trained model is evaluated on an independent validation data set to determine the effect of each parameter combination. Cross-validation is used or a portion of the data is reserved as a validation set to ensure the objectivity and reliability of the evaluation results.

[0136] Collect and record the performance characteristics of each parameter combination on the validation dataset, including accuracy and AUC (Area Under the Curve). Accuracy provides the proportion of correct classifications by the model, while AUC evaluates the performance of the model at different thresholds, especially in the case of unbalanced categories. Compare the accuracy and AUC values ​​of all parameter combinations and select the combination with the best overall performance.

[0137] Fine-tuning the model parameters can significantly improve the model's predictive accuracy and ensure that the model can reliably perform classification or prediction tasks. Evaluating the model on a validation dataset ensures that the selected parameters not only perform well on the training set, but also have good generalization capabilities on unseen data. Evaluation on the validation set helps identify overfitting parameter settings, thereby selecting parameter combinations that are stable on new data.

[0138] In some examples, the method for determining the second target parameter combination in the second parameter optimization mode includes:

[0139] Use Gaussian process to build a probability model between the parameter combination to be optimized and the model performance;

[0140] Calculate the performance loss of the above-mentioned parameter combination to be optimized in the current group according to the above-mentioned probability model;

[0141] The probability model is updated based on the performance loss and a new set of parameter combinations to be optimized, and the performance loss is calculated again based on the updated probability model;

[0142] Repeat the probability model update operation and the performance loss calculation operation to a preset number of times or until the performance parameter improvement is less than a preset threshold, and determine the corresponding target parameter combination at this time as the above-mentioned second target parameter combination to complete the second parameter optimization mode.

[0143] For example, Gaussian processes provide a powerful way to model and infer the relationship between parameters and performance. This method can not only predict performance, but also calculate the uncertainty of the prediction, which is crucial for subsequent parameter selection. The performance loss is calculated for the current set of parameter combinations to be optimized according to the above probability model. The performance loss is usually defined as the negative value of the objective function (for example, if the goal is to maximize accuracy, the loss is the negative accuracy). By calculating the performance loss, the optimization algorithm can determine which parameter regions are most likely to improve model performance.

[0144] The above probability model is updated based on the performance loss and a new set of parameter combinations to be optimized. The performance loss is calculated again based on the updated probability model. This step involves iteratively selecting new parameter combinations and updating the Gaussian process model with the newly obtained performance data. Through the iterative process, the model gradually learns and accurately maps the parameter space, thereby guiding parameter selection more effectively.

[0145] Repeat the probability model update operation and the performance loss calculation operation until the preset number of times or the performance parameter improvement is less than the preset threshold, and determine the corresponding parameter combination at this time as the second target parameter combination. Continue to improve the parameter combination until the performance improvement is no longer significant, ensuring that the found parameter combination is close to the global optimum.

[0146] In some examples, determining the optimal target parameter combination according to the first target parameter combination and the second target parameter combination includes:

[0147] Get the package promotion target;

[0148] When the package promotion target is the fast popular target, the first target parameter combination is determined as the optimal target parameter combination;

[0149] When the package promotion target is the maximum market target, the second target parameter combination is used to determine the optimal target parameter combination.

[0150] For example, first define and clarify the package promotion goals, which include rapid popularity goals or maximum market goals. Different package promotion goals reflect different market strategies and business priorities, such as rapidly increasing the user base or covering as large a market share as possible.

[0151] When the package promotion target is a rapid popularity target, the first target parameter combination is selected as the optimal target parameter combination.

[0152] When the package promotion target is the maximum market target, the second target parameter combination is selected as the optimal target parameter combination.

[0153] Depending on the package promotion objectives, the parameter combination that best promotes the goal is selected. This selection is based on the performance evaluation of the previous two optimization modes and the adaptability analysis of specific goals.

[0154] For strategies that pursue the goal of rapid popularity, choosing the first target parameter combination may mean that the combination can quickly adapt to market changes and accelerate user acceptance and engagement. For strategies that pursue the goal of the largest market, the second target parameter combination may focus more on long-term performance and stability, which helps to achieve in-depth promotion in a wider market.

[0155] like Figure 2 As shown, the present application proposes a user profile generation system based on user behavior, including:

[0156] The first collecting unit 21 is used to collect package attention data of the target user on specific websites and applications, wherein the package attention data includes package type information, package browsing frequency information, package browsing cumulative duration information and package purchase history information;

[0157] A first acquisition unit 22, configured to acquire the package attention feature to be predicted based on the package attention data;

[0158] The second collecting unit 23 is used to collect the package usage data of the target user, wherein the package usage data includes data usage, call usage and communication records of frequent contacts;

[0159] A second acquisition unit 24 is used to acquire the usage characteristics of the package to be predicted based on the package usage data;

[0160] The third collecting unit 25 is used to collect the feedback evaluation data of the target user, wherein the feedback evaluation data includes scoring information, evaluation information and feedback cycle information;

[0161] A third acquisition unit 26, configured to acquire the feedback evaluation feature to be predicted based on the feedback evaluation data;

[0162] The fourth acquisition unit 27 is used to input the above-mentioned package attention features to be predicted, the above-mentioned package usage features to be predicted and the above-mentioned feedback evaluation features to be predicted into the above-mentioned user portrait generation model to obtain the user portrait corresponding to the above-mentioned target user, wherein the above-mentioned user portrait includes the target recommended package type and the target package recommendation time.

[0163] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0164] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0165] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0166] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0167] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0168] An embodiment of the present application also provides a computer program product, which includes computer software instructions. When the computer software instructions are executed on a processing device, the processing device executes the process of predicting the decomposition rate in alumina production in the corresponding embodiment.

[0169] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on the computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium can be any available medium that a computer can store or a data storage device such as a server or a data center that includes one or more available media integration. The available medium can be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid state drive (SSD)), etc.

[0170] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0171] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0172] 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 units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0173] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0174] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), disk or optical disk and other media that can store program codes.

[0175] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for generating a user profile based on user behavior, comprising: Collect package attention data of target users on specific websites and applications, wherein the package attention data includes package type information, package browsing frequency information, package browsing cumulative duration information and package purchase history information; Acquire the package attention feature to be predicted based on the package attention data; Collecting package usage data of the target user, wherein the package usage data includes data usage, call usage and communication records of frequent contacts; Acquire usage characteristics of the package to be predicted based on the package usage data; Collecting feedback evaluation data of the target user, wherein the feedback evaluation data includes rating information, evaluation information and feedback cycle information; Acquire feedback evaluation features to be predicted based on the feedback evaluation data; The package attention features to be predicted, the package usage features to be predicted and the feedback evaluation features to be predicted are input into the user portrait generation model to obtain a user portrait corresponding to the target user, wherein the user portrait includes a target recommended package type and a target package recommendation time.

2. The method for generating a user profile based on user behavior according to claim 1, characterized in that: The obtaining the package attention feature to be predicted based on the package attention data includes: Calculating the package preference characteristics of each package type by weighted summation method according to the package browsing frequency information and the package purchase history information; Extracting package attention characteristics of each package type by time-weighted average method according to the package browsing cumulative time information; The K-means algorithm is used to analyze the distribution characteristics of the user's browsing time for the package according to the package browsing frequency information and the package browsing cumulative duration information.

3. The method for generating a user profile based on user behavior according to claim 1, characterized in that: The obtaining the usage characteristics of the package to be predicted based on the package usage data includes: Calculating a usage level feature according to the data usage and the call usage; Extract break time usage characteristics and working time usage characteristics based on the frequent contact communication records.

4. The method for generating a user profile based on user behavior according to claim 1, characterized in that: The obtaining the feedback evaluation feature to be predicted based on the feedback evaluation data includes: Determine the package satisfaction characteristics through statistical analysis based on the rating information; Determine emotional attitude characteristics based on the evaluation information through natural language processing; The feedback timeliness characteristics are determined using a time series analysis method based on the feedback cycle information.

5. The method for generating a user profile based on user behavior according to claim 1, characterized in that: The specific steps of the method for constructing the user portrait generation model include: A first training subset is constructed based on historical package attention features, a second training subset is constructed based on historical package usage features, a first training subset is constructed based on historical feedback evaluation features, and a fourth training subset is constructed based on user portraits corresponding to historical target users; Constructing a first decision tree corresponding to the package attention feature, a second decision tree corresponding to the package usage feature, and a third decision tree corresponding to the feedback evaluation feature; Training the first decision tree using the first training subset; Training the second decision tree using the second training subset; Training the third decision tree using the third training subset; Additional decision trees are added incrementally based on the gradient boosting method to correct the prediction residuals; The fourth training subset, the first decision tree, the second decision tree, the third decision tree, the additional decision tree, a voting module and a weighted average module are integrated to form the user portrait generation model, wherein the voting module is used to determine the target recommended package type by a voting method according to the output information of all decision trees, and the weighted average module is used to determine the target package recommendation time by a weighted average method according to the output information of all decision trees.

6. The method for generating a user profile based on user behavior according to claim 5, characterized in that: The specific steps of the optimization method of the user portrait generation model include: The depth and learning rate of the decision tree are determined as parameters to be optimized; Calculate the first target parameter combination under the first parameter optimization mode and the second target parameter combination under the second parameter optimization mode for all parameter combinations to be optimized; An optimal target parameter combination is determined according to the first target parameter combination or the second target parameter combination.

7. The method for generating a user profile based on user behavior according to claim 6, characterized in that: The method for determining the first target parameter combination in the first parameter optimization mode includes: Performing model training and validation operations on each set of parameter combinations to be optimized to obtain performance characteristics of the validation data set, wherein the performance characteristics include accuracy and AUC; A first target parameter combination is selected based on the performance characteristics to complete a first parameter optimization mode.

8. The method for generating a user profile based on user behavior according to claim 7, characterized in that: The method for determining the second target parameter combination in the second parameter optimization mode includes: Use Gaussian process to build a probability model between the parameter combination to be optimized and the model performance; Calculating the performance loss of the parameter combination to be optimized in the current group according to the probability model; updating the probability model based on the performance loss and a new set of parameter combinations to be optimized, and calculating the performance loss again based on the updated probability model; Repeat the probability model update operation and the performance loss calculation operation to a preset number of times or until the performance parameter improvement is less than a preset threshold, and determine the corresponding target parameter combination at this time as the second target parameter combination to complete the second parameter optimization mode.

9. The method for generating a user profile based on user behavior according to claim 8, characterized in that: The determining the optimal target parameter combination according to the first target parameter combination and the second target parameter combination includes: Get the package promotion target; When the package promotion target is the fast popular target, determining the first target parameter combination as the optimal target parameter combination; When the package promotion target is the maximum market target, the second target parameter combination is determined to be the optimal target parameter combination.

10. A user profile generation system based on user behavior, characterized in that: include: A first collecting unit is used to collect package attention data of target users on specific websites and applications, wherein the package attention data includes package type information, package browsing frequency information, package browsing cumulative duration information and package purchase history information; A first acquisition unit, configured to acquire the package attention feature to be predicted based on the package attention data; A second collecting unit is used to collect package usage data of the target user, wherein the package usage data includes data usage, call usage and communication records of frequent contacts; A second acquisition unit, configured to acquire the package usage characteristics to be predicted based on the package usage data; A third collecting unit, configured to collect feedback evaluation data of the target user, wherein the feedback evaluation data includes rating information, evaluation information and feedback cycle information; A third acquisition unit, configured to acquire the feedback evaluation feature to be predicted based on the feedback evaluation data; The fourth acquisition unit is used to input the package attention features to be predicted, the package usage features to be predicted and the feedback evaluation features to be predicted into the user portrait generation model to obtain the user portrait corresponding to the target user, wherein the user portrait includes the target recommended package type and the target package recommended time.

Images