Personalized service intelligent recommendation method and system based on data analysis
By building user portraits and material feature libraries, using high-thermal models and similar recall models, combining user behavior and time dimension weights, personalized recommendation results are generated, and the problem of insufficient capture of user interest changes in the existing technology is solved, and efficient and accurate intelligent recommendations are achieved.
Patent Information
- Application Number
- CN202510311708.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-17
- Publication Date
- 2025-07-11
AI Technical Summary
Existing personalized recommendation technologies are difficult to capture changes in user interest in real time when processing large-scale and multi-dimensional data, and are easily affected by cold start and data sparsity issues. Traditional methods ignore dynamic changes in users' short-term interests, resulting in insufficient accuracy and real-timeness of recommendation results. Overexposure of popular content affects the diversity and fairness of recommendations.
Data is obtained through the user behavior collection system, preprocessing and feature extraction, and user portraits and material feature libraries are built. High-thermal models and similar recall models are trained using big data analysis and artificial intelligence, behavior and time dimension weights are set, combination weights are calculated and scores are normalized, personalized recommendation results are generated, and model parameters are adjusted based on user feedback.
It improves the accuracy and adaptability of recommendations, ensures that the recommended content meets current hot trends, improves the accuracy and user experience of personalized recommendations, dynamically adapts to user needs, avoids over-exposure of popular content, and meets users' long-tail needs.
Smart Images
Figure CN120296244A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of intelligent recommendation, and particularly to a personalized service intelligent recommendation method and system based on data analysis. Background Art
[0002] With the rapid development of Internet technology and big data analysis, personalized recommendation systems have been widely applied in multiple fields such as e-commerce, online education, intelligent information push, social media, etc.; traditional recommendation systems mainly rely on methods such as collaborative filtering and content-based filtering, which can improve the user experience and reduce the problem of information overload to a certain extent; however, with the continuous enrichment and complexity of user behavior data, traditional recommendation methods have certain limitations in processing large-scale and multi-dimensional data and are difficult to capture the changing interests of users in real time; at the same time, the progress of artificial intelligence and deep learning technologies in recent years has promoted the development of recommendation systems, enabling them to analyze user preferences more accurately and make personalized recommendations; the combination of high-popularity recommendation models and similar recall models is becoming a mainstream trend, especially having significant advantages in improving the relevance of recommendation results and user satisfaction.
[0003] Existing personalized recommendation technologies still face many challenges; firstly, the accuracy of the recommendation system depends on the in-depth analysis of user behavior, but there is room for optimization in aspects such as setting behavior weights, considering time factors, and data normalization processing in existing methods, resulting in the recommendation results being susceptible to cold start problems and data sparsity problems; secondly, traditional recommendation algorithms often ignore the dynamic changes in users' short-term interests and overly rely on long-term behavior data, thus performing poorly in terms of real-time performance; in addition, some recommendation systems do not finely distinguish the contribution degrees of different behaviors when calculating high-popularity content, which easily leads to excessive exposure of popular content and affects the diversity and fairness of recommendations; therefore, how to improve the personalized accuracy and real-time performance of the recommendation system by optimizing the processing method of user behavior data, combining big data analysis and artificial intelligence algorithms is an urgent problem to be solved in the current intelligent recommendation field. Summary of the Invention
[0004] The purpose of the present invention is to provide a personalized service intelligent recommendation method and system based on data analysis to solve the problems raised in the above background art.
[0005] To solve the above technical problems, the present invention provides the following technical solutions:
[0006] A personalized service intelligent recommendation method based on data analysis, the method comprising the following steps: Step S1: Obtain user behavior data through a user behavior collection system, and perform preprocessing and feature extraction on the user behavior data; Step S2: Based on the preprocessed and feature-extracted user behavior data, construct a user portrait and a material feature library; Step S3: Use big data analysis and artificial intelligence technologies to train a recommendation algorithm model, the recommendation algorithm model including a high-heat model and a similar recall model; Step S4: Generate personalized recommendation results based on the user's real-time behavior and historical data; adjust the model parameters according to user feedback.
[0007] As a preferred solution of the personalized service intelligent recommendation method based on data analysis according to the present invention, the user behavior data includes: user login frequency, user access to services, user access to functions, user access frequency, and recommended material data, and the recommended material data includes material category data, content description data, and release time data.
[0008] As a preferred solution of the personalized service intelligent recommendation method based on data analysis according to the present invention, the high-heat model algorithm follows the following rules:
[0009] Set weights for different behavior and time dimensions, the behavior weights including click behavior weight, browse behavior weight, favorite behavior weight, and share behavior weight, and the time dimension weights including weekly time dimension weight and monthly time dimension weight;
[0010] According to the behavior weights and the time dimension weights, calculate the combined weights of different behaviors in different time periods, specifically: If the click behavior is within the weekly time dimension, the combined weight is the product of the click behavior weight and the weekly time dimension weight; If the click behavior is within the monthly time dimension, the combined weight is the product of the click behavior weight and the monthly time dimension weight; If the browse behavior is within the weekly time dimension, the combined weight is the product of the browse behavior weight and the weekly time dimension weight; If the browse behavior is within the monthly time dimension, the combined weight is the product of the browse behavior weight and the monthly time dimension weight; If the favorite behavior is within the weekly time dimension, the combined weight is the product of the favorite behavior weight and the weekly time dimension weight; If the favorite behavior is within the monthly time dimension, the combined weight is the product of the favorite behavior weight and the monthly time dimension weight; If the share behavior is within the weekly time dimension, the combined weight is the product of the share behavior weight and the weekly time dimension weight; If the share behavior is within the monthly time dimension, the combined weight is the product of the share behavior weight and the monthly time dimension weight;
[0011] Calculate the score of the content through a formula, and perform normalization processing on the calculated score, the score being equal to the sum of the combined weights of each behavior in different time periods multiplied by the occurrence times of the behavior in the corresponding time period.
[0012] As a preferred solution of the intelligent recommendation method for personalized services based on data analysis according to the present invention, the similar recall model algorithm follows the following rules:
[0013] Set weights for different behaviors and time dimensions. The behavior weights include click behavior weight, browsing behavior weight, favorite behavior weight, and sharing behavior weight. The time dimension weights include weekly time dimension weight and daily time dimension weight;
[0014] According to the behavior weights and the time dimension weights, calculate the combined weights of different behaviors in different time periods. Specifically: If the click behavior is within the weekly time dimension, the combined weight is the product of the click behavior weight and the weekly time dimension weight; if the click behavior is within the daily time dimension, the combined weight is the product of the click behavior weight and the daily time dimension weight; if the browsing behavior is within the weekly time dimension, the combined weight is the product of the browsing behavior weight and the weekly time dimension weight; if the browsing behavior is within the daily time dimension, the combined weight is the product of the browsing behavior weight and the daily time dimension weight; if the favorite behavior is within the weekly time dimension, the combined weight is the product of the favorite behavior weight and the weekly time dimension weight; if the favorite behavior is within the daily time dimension, the combined weight is the product of the favorite behavior weight and the daily time dimension weight; if the sharing behavior is within the weekly time dimension, the combined weight is the product of the sharing behavior weight and the weekly time dimension weight; if the sharing behavior is within the daily time dimension, the combined weight is the product of the sharing behavior weight and the daily time dimension weight;
[0015] Obtain the behavior data of the user for specific content within a specified number of days. The behavior data includes the number of clicks, the number of views, the number of favorites, and the number of shares;
[0016] Calculate the score of the content through a formula and normalize the calculated score. The score is equal to the sum of the products of the combined weights of each behavior in different time periods and the occurrence times of that behavior in the corresponding time periods;
[0017] Obtain the list of the user's recent preferred materials and their scores through a corresponding function, and sort them from high to low; calculate the similarity scores between all materials and the user's preferred materials, sort them from high to low, and select the top N materials as the recall results, where N is a pre-set value.
[0018] An intelligent recommendation system for personalized services based on data analysis. This system includes: a data collection and preprocessing module, a feature engineering module, a model training module, and a model deployment and analysis optimization module.
[0019] The data collection and preprocessing module: Obtain user behavior data through a user behavior collection system, and perform preprocessing and feature extraction on the user behavior data.
[0020] The feature engineering module: Based on the user behavior data after preprocessing and feature extraction, construct a user profile and a material feature library.
[0021] The model training module: Utilize big data analysis and artificial intelligence technologies to train a recommendation algorithm model, and the recommendation algorithm model includes a high-heat model and a similar recall model.
[0022] The model deployment and analysis optimization module: Generate personalized recommendation results based on the user's real-time behavior and historical data; adjust the model parameters according to user feedback.
[0023] Furthermore, the data collection and preprocessing module includes a data collection and preprocessing unit.
[0024] The data collection and preprocessing unit: Obtain user behavior data through a user behavior collection system, and perform preprocessing and feature extraction on the user behavior data; the user behavior data includes user login frequency, user access to services, user access to functions, user access frequency, and recommended material data, and the recommended material data includes material category data, content description data, and release time data.
[0025] Furthermore, the model training module includes a high-heat model training unit and a similar recall model training unit.
[0026] The high-heat model training unit: Set weights for different behavior and time dimensions. The behavior weights include click behavior weight, browse behavior weight, favorite behavior weight, and share behavior weight, and the time dimension weights include weekly time dimension weight and monthly time dimension weight; according to the behavior weights and the time dimension weights, calculate the combined weights of different behaviors in different time periods; calculate the score of the content through a formula, and perform normalization processing on the calculated score. The score is equal to the sum of the product of the combined weights of each behavior in different time periods and the occurrence times of that behavior in the corresponding time period.
[0027] The similar recall model training unit: sets weights for different behaviors and time dimensions. The behavior weights include click behavior weight, browsing behavior weight, favorite behavior weight, and sharing behavior weight. The time dimension weights include weekly time dimension weight and daily time dimension weight. According to the behavior weights and the time dimension weights, calculates the combined weights of different behaviors in different time periods. Obtains the behavior data of users for specific content within a specified number of days, where the behavior data includes the number of clicks, the number of views, the number of favorites, and the number of shares. Calculates the score of the content through a formula and normalizes the calculated score. The score is equal to the sum of the products of the combined weights of each behavior in different time periods and the occurrence times of that behavior in the corresponding time periods. Obtains the list of the user's recent preferred materials and their scores through a corresponding function, and sorts them from high to low. Calculates the similarity scores between all materials and the user's preferred materials, sorts them from high to low, and selects the top N materials as the recall results, where N is a pre-set value.
[0028] Further, the model deployment and analysis optimization module includes a model deployment and analysis optimization unit.
[0029] The model deployment and analysis optimization unit: applies the trained recommendation algorithm model to the recommendation system, and provides personalized recommendation services according to the real-time behaviors and interest changes of users. Optimizes the recommendation algorithm model according to user feedback and business development needs.
[0030] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: In the personalized service intelligent recommendation method and system based on data analysis provided by the present invention, user behavior data is obtained through the user behavior collection system, and the data is preprocessed and feature-extracted, laying a foundation for subsequent analysis. Based on the processed user behavior data, a user portrait and a material feature library are constructed, thereby forming the key features of the user preference model and the recommended content. The recommendation algorithm model is trained using big data analysis and artificial intelligence technologies, including a popularity model and a similar recall model, to improve the accuracy and adaptability of recommendations. The popularity model ensures that the recommended content conforms to the current hot trends by setting different behavior and time dimension weights, calculating the combined weights of different behaviors in different time periods, and normalizing the scores. The similar recall model improves the accuracy of personalized recommendations by calculating the similarities of all materials based on the list of the user's recent preferred materials and their scores, and selecting the top N materials as the recall results according to the scores. Based on the user's real-time behaviors and historical data, personalized recommendation results are generated, and the model parameters are adjusted according to user feedback, enabling the recommendation system to continuously optimize and adapt to user needs. Through the present invention, not only can more personalized recommendations that meet the user's interests be provided, improving the user experience and content matching degree, but also the dynamic adaptation ability of the recommendation system can be enhanced, thereby achieving more efficient and accurate intelligent recommendations. Brief Description of the Drawings
[0031] The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention.
[0032] Figure 1 is a schematic diagram of the steps of an intelligent recommendation method for personalized services based on data analysis according to the present invention;
[0033] Figure 2 is a schematic diagram of the structure of an intelligent recommendation system for personalized services based on data analysis according to the present invention. Detailed Embodiments
[0034] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.
[0035] Please refer to Figure 1 , in the first embodiment: Provide an intelligent recommendation method for personalized services based on data analysis, and the method includes the following steps:
[0036] Step S1: Obtain user behavior data through a user behavior collection system, and perform preprocessing and feature extraction on the user behavior data.
[0037] Specifically, the user behavior data includes: user login frequency, user access to services, user access to functions, user access frequency, and recommended material data. The recommended material data includes material category data, content description data, and release time data.
[0038] In the present invention, the accuracy of the recommendation system is often limited by data noise, redundant information, and data sparsity problems. Through preprocessing and feature extraction, invalid data can be effectively cleaned, and different feature scales can be normalized, so that subsequent recommendation algorithms can be trained based on high-quality data, improving the stability and accuracy of the recommendation effect.
[0039] Step S2: Based on the user behavior data after preprocessing and feature extraction, construct a user portrait and a material feature library.
[0040] In the present invention, traditional recommendation systems usually rely on simple behavior analysis and lack a deep understanding of user interests. Through user profile modeling, the present invention can depict users' long-term and short-term interests in a fine-grained manner, thereby improving the accuracy of recommendations. In addition, the establishment of a material feature library enables the recommendation system to make more accurate matches in the content dimension and avoid the drawbacks of generalized recommendations.
[0041] Step S3: Use big data analysis and artificial intelligence technologies to train a recommendation algorithm model, where the recommendation algorithm model includes a popularity model and a similarity recall model.
[0042] The popularity model algorithm follows the following rules:
[0043] Set weights for different behavior and time dimensions. The behavior weights include click behavior weight, browse behavior weight, favorite behavior weight, and share behavior weight. The time dimension weights include weekly time dimension weight and monthly time dimension weight.
[0044] According to the behavior weights and the time dimension weights, calculate the combined weights of different behaviors in different time periods. Specifically: If the click behavior is within the weekly time dimension, the combined weight is the product of the click behavior weight and the weekly time dimension weight; if the click behavior is within the monthly time dimension, the combined weight is the product of the click behavior weight and the monthly time dimension weight; if the browse behavior is within the weekly time dimension, the combined weight is the product of the browse behavior weight and the weekly time dimension weight; if the browse behavior is within the monthly time dimension, the combined weight is the product of the browse behavior weight and the monthly time dimension weight; if the favorite behavior is within the weekly time dimension, the combined weight is the product of the favorite behavior weight and the weekly time dimension weight; if the favorite behavior is within the monthly time dimension, the combined weight is the product of the favorite behavior weight and the monthly time dimension weight; if the share behavior is within the weekly time dimension, the combined weight is the product of the share behavior weight and the weekly time dimension weight; if the share behavior is within the monthly time dimension, the combined weight is the product of the share behavior weight and the monthly time dimension weight.
[0045] Calculate the score of the content through a formula and normalize the calculated score. The score is equal to the sum of the combined weights of each behavior in different time periods multiplied by the occurrence times of that behavior in the corresponding time period.
[0046] The similarity recall model algorithm follows the following rules:
[0047] Set weights for different behavior and time dimensions. The behavior weights include click behavior weight, browse behavior weight, favorite behavior weight, and share behavior weight. The time dimension weights include weekly time dimension weight and daily time dimension weight.
[0048] Calculate the combined weights of different behaviors in different time periods according to the behavior weights and the time dimension weights. Specifically: If the click behavior is within the weekly time dimension, the combined weight is the product of the click behavior weight and the weekly time dimension weight; if the click behavior is within the daily time dimension, the combined weight is the product of the click behavior weight and the daily time dimension weight; if the browse behavior is within the weekly time dimension, the combined weight is the product of the browse behavior weight and the weekly time dimension weight; if the browse behavior is within the daily time dimension, the combined weight is the product of the browse behavior weight and the daily time dimension weight; if the favorite behavior is within the weekly time dimension, the combined weight is the product of the favorite behavior weight and the weekly time dimension weight; if the favorite behavior is within the daily time dimension, the combined weight is the product of the favorite behavior weight and the daily time dimension weight; if the share behavior is within the weekly time dimension, the combined weight is the product of the share behavior weight and the weekly time dimension weight; if the share behavior is within the daily time dimension, the combined weight is the product of the share behavior weight and the daily time dimension weight.
[0049] Obtain the behavior data of the user for specific content within the specified number of days. The behavior data includes the number of clicks, the number of views, the number of favorites, and the number of shares.
[0050] Calculate the score of the content through a formula and normalize the calculated score. The score is equal to the sum of the products of the combined weights of each behavior in different time periods and the occurrence times of that behavior within the corresponding time period.
[0051] Obtain the list of the user's recent preferred materials and their scores through the corresponding function, and sort them from high to low; calculate the similarity scores between all materials and the user's preferred materials, sort them from high to low, and select the top N materials as the recall results, where N is a preset value.
[0052] In the present invention, the common problems of the existing recommendation systems include the overload of popular content (resulting in difficult recommendation of unpopular content) and insufficient coverage of short-term interests (unable to meet the long-tail needs of users). The high-heat model uses a time-weighted method to make the content that has received attention recently occupy a larger proportion in the recommendation, while the similar recall model ensures that the user's historical interests are not completely ignored. By calculating the combined weights of the behavior weights and the time dimension weights, the present invention further refines the influence of different behaviors on the recommendation results, improving the timeliness and accuracy of the recommendation.
[0053] Step S4: Generate personalized recommendation results based on the user's real-time behavior and historical data; adjust the model parameters according to the user feedback.
[0054] Specifically, apply the trained recommendation algorithm model to the recommendation system, provide personalized recommendation services for users according to the real-time behavior and interest changes of the users; optimize the recommendation algorithm model according to the user feedback and the business development needs.
[0055] In the present invention, the traditional recommendation system generally has the problem of model staticization, that is, it lacks the ability of dynamic adjustment after training, resulting in the difficulty of adapting the recommendation results to the changes in user interests. The present invention updates the recommendation strategy through the real-time behavior of users and optimizes the model parameters in combination with user feedback, enabling the recommendation system to adjust adaptively and improve the long-term recommendation effect.
[0056] Please refer to Figure 2 , in the second embodiment: A personalized service intelligent recommendation system based on data analysis is provided, and the system includes: a data collection and preprocessing module, a feature engineering module, a model training module, and a model deployment and analysis optimization module.
[0057] The data collection and preprocessing module: obtains user behavior data through a user behavior acquisition system, and preprocesses and extracts features from the user behavior data.
[0058] The feature engineering module: constructs a user portrait and a material feature library based on the user behavior data after preprocessing and feature extraction.
[0059] The model training module: uses big data analysis and artificial intelligence technologies to train a recommendation algorithm model, and the recommendation algorithm model includes a high-heat model and a similar recall model.
[0060] The model deployment and analysis optimization module: generates personalized recommendation results based on the real-time behavior and historical data of users; adjusts the model parameters according to user feedback.
[0061] Further, the data collection and preprocessing module includes a data collection and preprocessing module unit.
[0062] The data collection and preprocessing module unit: obtains user behavior data through a user behavior acquisition system, and preprocesses and extracts features from the user behavior data; the user behavior data includes user login frequency, user access to services, user access to functions, user access frequency, and recommended material data, and the recommended material data includes material category data, content description data, and release time data.
[0063] Further, the model training module includes a high-heat model training unit and a similar recall model training unit.
[0064] The high-temperature model training unit: sets weights for different behavior and time dimensions. The behavior weights include click behavior weight, browsing behavior weight, favorite behavior weight, and sharing behavior weight. The time dimension weights include weekly time dimension weight and monthly time dimension weight. According to the behavior weights and the time dimension weights, calculates the combined weights of different behaviors in different time periods. Calculates the score of the content through a formula and normalizes the calculated score. The score is equal to the sum of the products of the combined weights of each behavior in different time periods and the occurrence times of the behavior in the corresponding time periods.
[0065] The similar recall model training unit: sets weights for different behavior and time dimensions. The behavior weights include click behavior weight, browsing behavior weight, favorite behavior weight, and sharing behavior weight. The time dimension weights include weekly time dimension weight and daily time dimension weight. According to the behavior weights and the time dimension weights, calculates the combined weights of different behaviors in different time periods. Obtains the behavior data of the user for specific content within the specified number of days. The behavior data includes click times, browsing times, favorite times, and sharing times. Calculates the score of the content through a formula and normalizes the calculated score. The score is equal to the sum of the products of the combined weights of each behavior in different time periods and the occurrence times of the behavior in the corresponding time periods. Obtains the list of the user's recent preferred materials and their scores through a corresponding function and sorts them from high to low. Calculates the similarity scores between all materials and the user's preferred materials, sorts them from high to low, and selects the top N materials as the recall results, where N is a preset value.
[0066] Furthermore, the model deployment and analysis optimization module includes a model deployment and analysis optimization unit.
[0067] The model deployment and analysis optimization unit: applies the trained recommendation algorithm model to the recommendation system, provides personalized recommendation services according to the real-time behavior and interest changes of the user, and optimizes the recommendation algorithm model according to the user feedback and the business development needs.
[0068] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variation thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device.
[0069] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.
Claims
1. An intelligent recommendation method for personalized services based on data analysis, characterized in that, The method includes the following steps: Step S1: Obtain user behavior data through a user behavior collection system, and perform preprocessing and feature extraction on the user behavior data; Step S2: Based on the preprocessed and feature-extracted user behavior data, construct a user portrait and a material feature library; Step S3: Use big data analysis and artificial intelligence technologies to train a recommendation algorithm model, where the recommendation algorithm model includes a high-heat model and a similar recall model; Step S4: Generate personalized recommendation results based on the user's real-time behavior and historical data; adjust the model parameters according to user feedback.
2. The personalized service intelligent recommendation method based on data analysis according to claim 1, characterized in that, The user behavior data includes: user login frequency, user access to services, user access to functions, user access frequency, and recommended material data, and the recommended material data includes material category data, content description data, and release time data.
3. An intelligent recommendation method for personalized services based on data analysis according to claim 2, characterized in that, The high-heat model algorithm follows the following rules: Set weights for different behavior and time dimensions. The behavior weights include click behavior weight, browse behavior weight, favorite behavior weight, and share behavior weight, and the time dimension weights include weekly time dimension weight and monthly time dimension weight; According to the behavior weights and the time dimension weights, calculate the combined weights of different behaviors in different time periods. Specifically: If the click behavior is within the weekly time dimension, the combined weight is the product of the click behavior weight and the weekly time dimension weight; if the click behavior is within the monthly time dimension, the combined weight is the product of the click behavior weight and the monthly time dimension weight; if the browse behavior is within the weekly time dimension, the combined weight is the product of the browse behavior weight and the weekly time dimension weight; if the browse behavior is within the monthly time dimension, the combined weight is the product of the browse behavior weight and the monthly time dimension weight; if the favorite behavior is within the weekly time dimension, the combined weight is the product of the favorite behavior weight and the weekly time dimension weight; if the favorite behavior is within the monthly time dimension, the combined weight is the product of the favorite behavior weight and the monthly time dimension weight; if the share behavior is within the weekly time dimension, the combined weight is the product of the share behavior weight and the weekly time dimension weight; if the share behavior is within the monthly time dimension, the combined weight is the product of the share behavior weight and the monthly time dimension weight; Calculate the score of the content through a formula and perform normalization processing on the calculated score. The score is equal to the sum of the combined weights of each behavior in different time periods multiplied by the occurrence times of the behavior in the corresponding time period.
4. An intelligent recommendation method for personalized services based on data analysis according to claim 3, characterized in that, The similar recall model algorithm follows the following rules: Set weights for different behavior and time dimensions. The behavior weights include click behavior weight, browse behavior weight, favorite behavior weight, and share behavior weight, and the time dimension weights include weekly time dimension weight and daily time dimension weight; Calculate the combined weights of different behaviors in different time periods according to the behavior weights and the time dimension weights. Specifically: If the click behavior is within the weekly time dimension, the combined weight is the product of the click behavior weight and the weekly time dimension weight; if the click behavior is within the daily time dimension, the combined weight is the product of the click behavior weight and the daily time dimension weight; if the browsing behavior is within the weekly time dimension, the combined weight is the product of the browsing behavior weight and the weekly time dimension weight; if the browsing behavior is within the daily time dimension, the combined weight is the product of the browsing behavior weight and the daily time dimension weight; if the collection behavior is within the weekly time dimension, the combined weight is the product of the collection behavior weight and the weekly time dimension weight; if the collection behavior is within the daily time dimension, the combined weight is the product of the collection behavior weight and the daily time dimension weight; if the sharing behavior is within the weekly time dimension, the combined weight is the product of the sharing behavior weight and the weekly time dimension weight; if the sharing behavior is within the daily time dimension, the combined weight is the product of the sharing behavior weight and the daily time dimension weight; Obtain the behavior data of the user for specific content within the specified number of days. The behavior data includes the number of clicks, the number of views, the number of collections, and the number of shares; Calculate the score of the content through a formula and normalize the calculated score. The score is equal to the sum of the products of the combined weights of each behavior in different time periods and the occurrence times of that behavior in the corresponding time period; Obtain the list of the user's recent preferred materials and their scores through the corresponding function, and sort them from high to low; calculate the similarity scores between all materials and the user's preferred materials, sort them from high to low, and select the top N materials as the recall results, where N is a pre-set value.
5. A personalized service intelligent recommendation system based on data analysis, which executes a personalized service intelligent recommendation method based on data analysis according to any one of claims 1-4, characterized in that, The system includes: a data collection and preprocessing module, a feature engineering module, a model training module, and a model deployment and analysis optimization module; The data collection and preprocessing module: Obtain user behavior data through the user behavior acquisition system, and perform preprocessing and feature extraction on the user behavior data; The feature engineering module: Based on the user behavior data after preprocessing and feature extraction, construct a user portrait and a material feature library; The model training module: Use big data analysis and artificial intelligence technologies to train a recommendation algorithm model. The recommendation algorithm model includes a high-heat model and a similarity recall model; The model deployment and analysis optimization module: Generate personalized recommendation results based on the user's real-time behavior and historical data; adjust the model parameters according to the user feedback.
6. An intelligent recommendation system for personalized services based on data analysis according to claim 5, characterized in that: The data collection and preprocessing module includes a data collection and preprocessing unit; The data collection and preprocessing unit: Obtain user behavior data through the user behavior acquisition system, and perform preprocessing and feature extraction on the user behavior data; the user behavior data includes the user login frequency, the user access service, the user access function, the user access frequency, and the recommended material data. The recommended material data includes the material category data, the content description data, and the release time data.
7. An intelligent recommendation system for personalized services based on data analysis according to claim 6, characterized in that: The model training module includes a high-heat model training unit and a similarity recall model training unit; The high-heat model training unit: sets weights for different behavior and time dimensions. The behavior weights include click behavior weight, browsing behavior weight, collection behavior weight, and sharing behavior weight. The time dimension weights include weekly time dimension weight and monthly time dimension weight. According to the behavior weights and the time dimension weights, calculates the combined weights of different behaviors in different time periods. Calculates the score of the content through a formula and normalizes the calculated score. The score is equal to the sum of the products of the combined weights of each behavior in different time periods and the occurrence times of that behavior in the corresponding time period. The similar recall model training unit: sets weights for different behavior and time dimensions. The behavior weights include click behavior weight, browsing behavior weight, collection behavior weight, and sharing behavior weight. The time dimension weights include weekly time dimension weight and daily time dimension weight. According to the behavior weights and the time dimension weights, calculates the combined weights of different behaviors in different time periods. Obtains the behavior data of the user for specific content within the specified number of days. The behavior data includes click times, browsing times, collection times, and sharing times. Calculates the score of the content through a formula and normalizes the calculated score. The score is equal to the sum of the products of the combined weights of each behavior in different time periods and the occurrence times of that behavior in the corresponding time period. Obtains the list of the user's recent preferred materials and their scores through a corresponding function and sorts them from high to low. Calculates the similarity scores between all materials and the user's preferred materials, sorts them from high to low, and selects the top N materials as the recall results, where N is a pre-set value.
8. An intelligent recommendation system for personalized services based on data analysis according to claim 7, characterized in that: The model deployment and analysis optimization module includes a model deployment and analysis optimization unit. The model deployment and analysis optimization unit: applies the trained recommendation algorithm model to the recommendation system and provides personalized recommendation services according to the real-time behavior and interest changes of the user. Optimizes the recommendation algorithm model according to user feedback and business development requirements.