Personalized recommendation algorithm and system based on AI

Abstract

The invention provides an AI-based personalized recommendation algorithm and system. The system comprises a data collection module, a user portrait module, a recommendation engine module, a real-time feedback module, a data prejudice processing module and a cold start solution. According to the method, the accuracy and coverage rate of recommendation can be improved, diversified personalized requirements of users are met, interests and requirements of the users can be better matched, and the satisfaction degree of the users is improved; the problems of cold start and data prejudice can be solved, the robustness and stability of the system are enhanced, and personalized recommendation services can be quickly and accurately provided; the recommendation result can be updated in real time, a more flexible and intelligent recommendation service is provided for the user, the behavior change of the user can be reflected in the recommendation result in time, and the recommendation timeliness is ensured; by constructing the refined user portrait, the system can deeply understand the real demand of the user, more accurate and personalized recommendation service is provided, and the user experience is improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and particularly to an AI-based personalized recommendation algorithm and system. Background Art

[0002] In today's information age, Internet technology has shown an unprecedented rapid development trend. When facing a vast ocean of information resources, users often fall into a dilemma of being at a loss and find it difficult to quickly and efficiently obtain valuable information that truly meets their own needs. Against this background, personalized recommendation systems have emerged as the key weapon to solve the problem of information overload. Existing personalized recommendation algorithms mainly include various types such as collaborative filtering, content-based recommendation, and hybrid recommendation. However, these algorithms have exposed many obvious limitations in the actual application process. For example, they are inefficient in processing large-scale data sets, difficult to accurately capture the complex and changeable behavior patterns of users, and also appear powerless in dealing with cold start problems and data bias problems. The cold start problem is mainly reflected in the lack of sufficient historical data for new users or new projects, resulting in a significant decline in recommendation accuracy; while the data bias problem stems from the dominant position of popular projects in the system, making the recommendation results tend to popular projects and ignoring the value of long-tail projects. Therefore, there is an urgent need for a more efficient, accurate and personalized recommendation algorithm and system that can effectively solve the above problems to meet the growing personalized needs of users. Summary of the Invention

[0003] The purpose of the present invention is to provide an AI-based personalized recommendation algorithm and system, aiming to solve cold start and data bias problems, improve the recommendation effect, and meet the diverse needs of users.

[0004] To achieve the above invention purpose, the present invention provides the following technical solutions:

[0005] The present invention provides an AI-based personalized recommendation algorithm and system, including a data collection module, a user profile module, a recommendation engine module, a real-time feedback module, a data bias processing module, and a cold start solution.

[0006] Further, the recommendation engine module includes a collaborative filtering sub-module, a content-based recommendation sub-module, and a hybrid recommendation sub-module.

[0007] Further, the data bias processing module introduces a weight balancing mechanism.

[0008] Further, the cold start solution includes introducing content-based recommendation and a preference questionnaire at the time of user registration.

[0009] Further, the data collection module collects information such as users' browsing records, purchase histories, and social media activities on various platforms through API interfaces and log files;

[0010] The user profiling module uses a number of learning algorithms for feature extraction and model training to construct a refined user profile;

[0011] The collaborative filtering sub-module generates preliminary recommendation results by analyzing the similarity between users and the similarity between items.

[0012] Further, the hybrid recommendation sub-module organically integrates the collaborative filtering and content-based recommendation results;

[0013] The real-time feedback module dynamically adjusts and updates the recommendation algorithm and model by monitoring and analyzing the feedback information of users on the recommended content in real time.

[0014] Further, the data bias processing module down-weights popular items and up-weights unpopular items;

[0015] The cold start solution combines content-based recommendation with the preference questionnaire at the time of user registration.

[0016] An AI-based personalized recommendation method, which includes the following steps:

[0017] a) Collect information such as users' browsing records, purchase histories, and social media activities;

[0018] b) Construct a refined user profile, including features in multiple dimensions such as age, gender, interest preferences, and consumption habits;

[0019] c) Use deep learning technology to construct a neural network model, conduct multi-dimensional and in-depth analysis of user behavior, and generate a recommendation list;

[0020] d) Monitor and analyze the feedback information of users on the recommended content in real time, and dynamically adjust and update the recommendation algorithm and model;

[0021] e) Introduce a weight balancing mechanism to reduce the excessive influence of popular items on the recommendation results and improve the exposure of long-tail items;

[0022] f) Use content-based recommendation methods and the preference questionnaire at the time of user registration to solve the cold start problem.

[0023] Beneficial effects:

[0024] The present invention provides an AI-based personalized recommendation algorithm and system, including a data collection module, a user profile module, a recommendation engine module, a real-time feedback module, a data bias processing module, and a cold start solution. The present invention can improve the accuracy and coverage of recommendations, meet the diverse personalized needs of users, better match the interests and needs of users, and enhance user satisfaction; it can solve the cold start and data bias problems, enhance the robustness and stability of the system, and provide personalized recommendation services quickly and accurately; it can update the recommendation results in real time, provide more flexible and intelligent recommendation services for users, and the behavior changes of users can be reflected in the recommendation results in a timely manner to ensure the timeliness of recommendations; by constructing a refined user profile, the system can deeply understand the real needs of users, provide more accurate and personalized recommendation services, and enhance the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on the provided drawings.

[0026] Figure 1 It is the system framework diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the 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 of 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.

[0028] An AI-based personalized recommendation algorithm and system disclosed by the present invention includes a data collection module, a user profile module, a recommendation engine module, a real-time feedback module, a data bias processing module, and a cold start solution.

[0029] An AI-based personalized recommendation method, the method includes the following steps:

[0030] a) Collect information such as the browsing records, purchase history, and social media activities of users;

[0031] b) Construct a refined user profile, including features in multiple dimensions such as age, gender, interest preferences, and consumption habits;

[0032] c) Adopt deep learning technology to build a neural network model, conduct multi-dimensional and in-depth analysis of user behavior, and generate a recommendation list;

[0033] d) Real-time monitor and analyze the feedback information of users on the recommended content, and dynamically adjust and update the recommendation algorithm and model;

[0034] e) Introduce a weight balancing mechanism to reduce the excessive influence of popular items on the recommendation results and increase the exposure of long-tail items;

[0035] f) Adopt a content-based recommendation method and a preference questionnaire at the time of user registration to solve the cold start problem.

[0036] To further optimize the above technical solution, the data collection module widely collects information such as browsing records, purchase history, and social media activities of users on various platforms through multiple methods such as API interfaces and log files. These data sources are rich and diverse, ensuring the comprehensiveness and diversity of the data, and providing sufficient data support for subsequent recommendation algorithms.

[0037] To further optimize the above technical solution, the user profiling module first conducts comprehensive cleaning and preprocessing on the collected data, removes invalid data and noise data, and ensures the accuracy and usability of the data. Then, machine learning algorithms such as clustering and classification are used for feature extraction and model training to build a refined user profile. The specific steps include data cleaning, data preprocessing, feature extraction, and model training, etc. Each link is carefully designed and optimized to ensure the accuracy and comprehensiveness of the user profile.

[0038] To further optimize the above technical solution, the recommendation engine module is one of the core components of the entire system. It includes a collaborative filtering sub-module, a content-based recommendation sub-module, and a hybrid recommendation sub-module. The collaborative filtering sub-module generates preliminary recommendation results by deeply analyzing the similarity between users and the similarity between items. The content-based recommendation sub-module generates content-based recommendation results by carefully analyzing the feature information of items, such as categories, tags, attributes, etc. The hybrid recommendation sub-module organically combines the collaborative filtering and content-based recommendation results, comprehensively considers various factors, and finally generates a more accurate and diverse recommendation list.

[0039] To further optimize the above technical solution, the real-time feedback module dynamically adjusts and updates the recommendation algorithm and model by real-time monitoring and analyzing the feedback information of users on the recommended content, such as click-through rate, browsing time, purchase behavior, etc. When users generate new feedback on the recommended content, the system can respond quickly and adjust the recommendation strategy in a timely manner to ensure the accuracy and timeliness of the recommendation results.

[0040] To further optimize the above technical solution, the data bias processing module effectively solves the data bias problem by introducing an innovative weight balancing mechanism. The specific methods include appropriately reducing the weight of popular items to lower their weight ratio in the recommendation results and avoid popular items overly occupying the recommendation resources. At the same time, weighted processing is performed on unpopular items to increase their exposure in the recommendation list, enabling more valuable long-tail items to be discovered and concerned by users.

[0041] To further optimize the above technical solution, the cold start solution includes two aspects: introducing content-based recommendation and a preference questionnaire when users register. Content-based recommendation generates initial recommendation results by deeply analyzing the preference information provided by users when registering, such as hobbies and favorite categories, providing preliminary personalized recommendation services for new users. The preference questionnaire comprehensively collects users' interest preferences by designing a series of elaborate questions, serving as the initial data support for the system to build user portraits and recommendation algorithms. The combination of these two methods can effectively solve the cold start problem and improve the recommendation accuracy for new users.

[0042] The present invention provides an AI-based personalized recommendation algorithm and system, which can improve the accuracy and coverage of recommendations, meet the diverse personalized needs of users, better match users' interests and needs, and enhance user satisfaction; it can solve the cold start and data bias problems, enhance the robustness and stability of the system, and quickly and accurately provide personalized recommendation services; it can update the recommendation results in real time, provide more flexible and intelligent recommendation services for users, and the changes in users' behaviors can be timely reflected in the recommendation results to ensure the timeliness of recommendations; by constructing a refined user portrait, the system can deeply understand users' real needs, provide more accurate and personalized recommendation services, and enhance the user experience.

[0043] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can still be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. AI-based personalized recommendation algorithm and system, characterized by: It includes data collection module, user portrait module, recommendation engine module, real-time feedback module, data bias processing module and cold start solution.

2. The AI-based personalized recommendation algorithm and system according to claim 1, characterized in that: The recommendation engine module includes a collaborative filtering submodule, a content-based recommendation submodule and a hybrid recommendation submodule.

3. The AI-based personalized recommendation algorithm and system according to claim 1 or 2, characterized in that: The data bias processing module introduces a weight balancing mechanism.

4. The AI-based personalized recommendation algorithm and system according to claim 1 or 2, characterized in that: The cold start solution includes the introduction of content-based recommendations and a preference questionnaire upon user registration.

5. The AI-based personalized recommendation algorithm and system according to any one of claims 1 to 4, characterized in that: The data collection module collects information such as users’ browsing history, purchase history, social media activities, etc. on various platforms through API interfaces and log files; The user portrait module uses several learning algorithms to perform feature extraction and model training to construct a refined user portrait; The collaborative filtering submodule generates preliminary recommendation results by analyzing the similarities between users and the similarities between items.

6. The AI-based personalized recommendation algorithm and system according to any one of claims 1 to 4, characterized in that: The hybrid recommendation submodule organically integrates collaborative filtering and content-based recommendation results; The real-time feedback module dynamically adjusts and updates the recommendation algorithm and model by real-time monitoring and analyzing the user's feedback information on the recommended content.

7. The AI-based personalized recommendation algorithm and system according to any one of claims 1 to 4, characterized in that: The data bias processing module performs down-weighting processing on popular items and weighting processing on unpopular items; The cold start solution is achieved by combining content-based recommendations with a preference questionnaire during user registration.

8. A personalized recommendation method based on AI, the method comprising the following steps: a) Collect information such as users’ browsing history, purchase history, social media activities, etc.; b) Build a refined user profile, including characteristics of multiple dimensions such as age, gender, interest preferences, and consumption habits; c) Use deep learning technology to build a neural network model, conduct multi-dimensional and in-depth analysis of user behavior, and generate a recommendation list; d) Real-time monitoring and analysis of user feedback on recommended content, and dynamic adjustment and update of recommendation algorithms and models; e) Introduce a weight balancing mechanism to reduce the excessive influence of popular projects on recommendation results and increase the exposure of long-tail projects; f) Use content-based recommendation methods and user preference questionnaires during registration to solve the cold start problem.

Images