Intelligent recommendation method and system based on online effect scoring and recall link fusion

Abstract

The invention discloses an intelligent recommendation method and system based on online effect score and recall link fusion. The method comprises the steps that multiple content recall links are set according to online provided content and content categories; collecting user data generated when the user accesses the content; calculating to obtain an online effect score and a weight parameter of each recall link according to the user data; and performing weighted summation according to the online effect score and the weight parameter of the recall link to obtain a recommendation score of the recall link, and by introducing a real-time scoring mechanism based on the online effect and fully exerting the advantages of different recall strategies, personalized, accurate and diversified recommendation results are realized.

Description

Technical Field

[0001] This invention relates to an intelligent recommendation method and system based on the fusion of online performance scoring and recall links, belonging to the field of content recommendation technology. Background Technology

[0002] In existing technologies, recommender systems mainly rely on three types of techniques to achieve content recommendation: collaborative filtering, content-based recommendation, and hybrid recommendation. These methods primarily filter and rank content based on users' historical behavior data, interest tags, and content features, with the following specific characteristics: Collaborative filtering: User-to-user collaborative filtering: This method calculates the similarity between users and uses historical interaction data of users with similar interests or behaviors as a basis for recommendations. Common methods used to measure user similarity include cosine similarity and Pearson correlation coefficient.

[0003] Item-to-item collaborative filtering: By analyzing user ratings or interaction records of different items, it recommends other items with similar characteristics to those of the user's historical preferences. This method relies on a matrix of associations between items.

[0004] Content-based recommendation: Recommendations are made by analyzing users' historical interactions with content (such as clicks, views, or purchases) and combining this with the characteristics of the content itself (such as text, tags, and images). Common methods include TF-IDF-based text feature extraction and content vector similarity calculation.

[0005] This type of method focuses on the degree of matching between user preferences and content features, and is suitable for content scenarios with rich structured features.

[0006] Hybrid Recommendation System: By combining collaborative filtering and content-based recommendation strategies, a more comprehensive recommendation effect can be achieved. Common methods include weighted fusion of the recommendation results from the two strategies and applying different strategies in stages. Hybrid recommendation systems attempt to combine the advantages of both to improve the accuracy and diversity of recommendations.

[0007] However, the aforementioned existing technologies have the following problems when implemented: The recall strategy is too simplistic and lacks flexibility. Existing recommender systems primarily rely on collaborative filtering, content-based recommendation, or hybrid recommendation strategies, typically employing only a few simple recall methods. While these methods are effective in certain scenarios, they lack sufficient flexibility when dealing with complex user behavior data, diverse user needs, and the multiple characteristics of the content itself. For example, collaborative filtering performs poorly in sparse data situations, while content-based recommendation can easily lead to overly simplistic recommendations lacking diversity and innovation.

[0008] User and content preferences are not calculated accurately enough: Traditional recommendation systems typically infer user preferences based on static interest tags, click history, or content similarity, ignoring the dynamic changes in user behavior. Because they lack dynamic and precise metrics for the uniqueness (differentiation coefficient) between users and content, they struggle to capture users' nuanced preferences for specific content types and cannot effectively assess the relative scarcity or popularity of content within the recommendation pool.

[0009] The shortcomings of recall link integration: Existing recall link fusion strategies are typically quite simple, lacking multi-dimensional fusion optimization mechanisms. Most systems simply perform weighted averaging or ranking of the results from different recall strategies, failing to dynamically adjust the effectiveness of each recall link. This approach often results in low-quality recommendation results, failing to maximize the satisfaction of user needs.

[0010] Traditional weighted fusion methods are typically based on fixed weights and lack real-time data feedback and adaptive adjustment capabilities. Therefore, this static weighting approach may introduce biases in certain situations and fail to fully utilize the various recall links. Summary of the Invention

[0011] The purpose of this invention is to overcome the shortcomings of the prior art and provide an intelligent recommendation method and system based on the integration of online performance scoring and recall links. By introducing a real-time scoring mechanism based on online performance and giving full play to the advantages of different recall strategies, personalized, accurate and diversified recommendation results can be achieved.

[0012] To achieve the above objectives, the present invention is implemented using the following technical solution: In a first aspect, the present invention provides an intelligent recommendation method based on the fusion of online performance scoring and recall links, comprising: Set up multiple content recall links based on the content provided online and the content categories; Collect user data generated when users access content; The online performance score and weight parameters for each content recall link are calculated based on user data. The recommendation score for the content recall link is obtained by weighting and summing the online performance scores and weight parameters of the content recall link.

[0013] Furthermore, the user data includes behavioral data generated by user access and click behavior within a preset time window, and content feature data obtained based on the attribute data of the accessed product.

[0014] Furthermore, the preset time window is 60 days.

[0015] Furthermore, the calculation of the online performance score and weight parameters for each content recall link based on user data includes: The user differentiation coefficient, which represents a user's preference for a certain type of content recall link, and the content differentiation coefficient, which represents the attention received by the content, are calculated based on user data. The online performance score and weight parameters for each content recall link are calculated based on the user differentiation coefficient and the content differentiation coefficient.

[0016] Furthermore, the user differentiation coefficient Content Differentiation Coefficient The mathematical expression is: ; ; In the formula, This represents the number of interactions between the i-th user and the k-th content retrieval link; This represents the total number of interactions between the i-th user and all content retrieval links; n is the total number of users, and m is the total number of content retrieval links; This represents the number of interactions of the k-th content retrieval link with the i-th user. This represents the total number of interactions among all users for the k-th content recall link.

[0017] Furthermore, the calculation of the online performance score and weight parameters for each content recall link based on the user differentiation coefficient and content differentiation coefficient includes: The online performance of the content recall process is scored based on user differentiation coefficient, content differentiation coefficient, and content click-through rate. Standardize the online performance scores to obtain standardized output; The weight parameters for each content recall link are calculated based on the online performance score.

[0018] Furthermore, the online performance evaluation of the content recall link based on the user differentiation coefficient, content differentiation coefficient, and content click-through rate is expressed as follows: ; In the formula, score represents the online performance rating, and r is the ranking of different content recall links after ranking based on content click-through rate. The ranking is sorted from highest to lowest based on the click-through rate of user behavior data.

[0019] Furthermore, the online performance score is standardized to obtain the standardized output `normalized_score`, expressed as:

[0020] In the formula, min(score) represents the lowest score among all online performance scores, and max(score) represents the highest score among all online performance scores.

[0021] Furthermore, the weight parameter (weight) for each content recall link is calculated based on the online performance score, and its expression is: ; The standardized scores from different recall strategies are weighted and summed according to the weight parameters to calculate the user's final recommendation score, weighted_avg_score, as expressed by: .

[0022] Secondly, the present invention provides an intelligent recommendation system based on the fusion of online performance scoring and recall links, comprising: The data collection module is used to collect users' historical behavior data, financial content information, and market trend data; The data preprocessing module is used to preprocess the collected data to obtain processed data; The multi-strategy recall module is used to implement multiple recall strategies based on the content recall links set according to the content tag categories, based on the processed data. The online performance scoring module is used to calculate the online performance score and weight parameters for each content recall link based on user data generated when users access content. The weighted fusion and recommendation ranking module is used to sum the online performance scores of different content recall links according to the weight parameters to calculate the recommendation score; The dynamic weight adjustment mechanism module is used to periodically adjust weight parameters based on user behavior feedback.

[0023] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: This invention provides an intelligent recommendation method and system based on the fusion of online performance scoring and recall links. By setting multiple content recall links according to the online content and content categories, a multi-strategy recall link fusion mechanism is formed, which can give full play to the advantages of different recall strategies to achieve personalized, accurate and diversified recommendation results.

[0024] This invention also collects user data generated when users access content; calculates the online performance score and weight parameters of each content recall link based on the user data; and performs a weighted summation based on the online performance score and weight parameters of the content recall link to obtain the recommendation score of the content recall link. Thus, by introducing a real-time scoring mechanism based on online performance, the weighted fusion strategy can be continuously optimized to ensure that the recommendation system can quickly respond to changes in user interests and adjust the recommended content in real time. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating an intelligent recommendation method based on the fusion of online performance scoring and recall links provided in an embodiment of the present invention. Detailed Implementation

[0026] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0027] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are used only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0028] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances. Example

[0029] Please see Figure 1 This embodiment introduces an intelligent recommendation method based on the fusion of online performance scoring and recall links, including: S100. Collect behavioral and content data generated during user visits.

[0030] Specifically, based on multi-day user access and click behavior data, an online performance score is calculated for content recommendations within the financial app. This score integrates user and product differentiation coefficients with content click-through rate ranking values ​​to measure the effectiveness of various recall strategies. This scoring method dynamically reflects user preferences and content popularity, thus providing a foundation for subsequent recommendation results.

[0031] First, user behavior data and product attribute data are obtained from multiple data sources. The behavior data includes user click and access behavior data (UV and CTR), and the product attribute data is content feature data.

[0032] In this embodiment, the time window for collecting behavioral and content data is set to 60 days, and parallel processing is performed using the Spark big data platform to improve computational efficiency. After cleaning, deduplication, and formatting, the data is used as input for subsequent online score calculations.

[0033] S200. Calculate and obtain the user differentiation coefficient based on behavioral data and content data. Content Differentiation Coefficient The content differentiation coefficient can also represent the content uniqueness rate.

[0034] Among them, the user differentiation coefficient The content differentiation coefficient represents a user's preference for a particular type of content retrieval path. This measure assesses the level of attention content receives within the recommendation pool, essentially reflecting the relative scarcity or popularity of content across the entire content retrieval chain. The calculation formula is as follows: ; ; In the formula, This represents the number of interactions (clicks) between the i-th user and the k-th content recall link. This represents the total number of interactions between the i-th user and all content retrieval links. n is the total number of users, and m is the total number of content retrieval links. This represents the number of interactions of the k-th content retrieval link with the i-th user. This represents the total number of interactions across all users in the k-th content recall link. Through fine-grained statistical analysis of behavioral and content data, the system can dynamically track user and product performance in the recommendation environment.

[0035] S300, Calculation of Online Performance Rating Online rating based on real-time behavioral data: Online performance rating not only considers the difference coefficient between users and content, but also combines click-through rate (CTR) and user interaction behavior to ensure that the recommended content can effectively stimulate user interest.

[0036] Rating smoothing: To avoid excessive fluctuations in ratings, the differentiation coefficient is smoothed during calculation. This process preserves the dynamic changes in ratings while reducing the impact of extreme values, thus ensuring recommendation stability.

[0037] Click-through rate (CTR) ranking weighted adjustment: The online performance scoring introduces the key variable of content click-through rate ranking to ensure that the best-performing content is recommended first among similar content, realizing natural competition among content and improving the priority ranking of recommendations.

[0038] The formula for calculating online ratings is: ; Here, "score" represents the online performance rating, and "r" represents the ranking of different content recall paths based on click-through rate. The ranking is based on click-through rate from highest to lowest according to user behavior data.

[0039] In the implementation process, data partitioning and computation, along with SQL optimization (database system optimization) and parallel computing techniques, significantly improve processing speed. The scoring results are ultimately stored in Parquet format, ensuring efficient data storage and fast retrieval.

[0040] Parquet is an open-source columnar storage file format designed for efficient storage and processing of large-scale data, offering advantages such as data compression, columnar storage, and language independence.

[0041] S300, Adaptive Weighted Fusion of Multi-Strategy Recall. To ensure the high quality and diversity of recommended content, a dynamic weight adjustment mechanism is added to the fusion of various recall strategies to achieve intelligent adaptive recommendation results.

[0042] Specifically, standardization and multi-dimensional weight allocation: Online scores for each content recall link are standardized to eliminate differences in scoring scales before weighting. The weight design fully considers the contributions of different strategies to user experience and content diversity, avoiding the excessive influence of a single strategy. The standardization formula is: ; In the formula, normalized_score represents the standardized output, min(score) represents the lowest score among all online performance scores, and max(score) represents the highest score among all online performance scores.

[0043] After standardization, the system calculates the weight for each recall type using the following formula: ; This weighting mechanism ensures that different recall types play a balanced role in the fusion process, avoiding excessive influence of a particular type of content recall link on the final recommendation results.

[0044] Weighted Summation and Adaptive Adjustment: The system generates a final recommendation score by weighting and summing the standardized scores from different content recall links. This weighting mechanism incorporates online scoring feedback from the recall strategy, dynamically adjusting the weight of each content recall link to meet users' evolving interests.

[0045] The system calculates the user's final recommendation score by weighting and summing the standardized scores from different recall strategies according to their respective weights. ; Through this weighted fusion strategy, the system can adjust according to the advantages of different recall types to achieve personalized ranking of recommendation results.

[0046] S400, generation and diversity optimization of personalized recommendation lists.

[0047] Based on the personalized ranking of the recommendation results, a final recommendation list is generated. In generating the final recommendation list, the system achieves a high balance between content diversity and interest relevance to ensure that the recommended content received by users matches their interests and preferences and possesses high discovery value.

[0048] It should be added that this invention can be efficiently deployed in recommendation systems of large-scale financial institutions. By leveraging distributed data processing architectures (such as Hadoop / Spark) for big data processing and through periodic batch processing and real-time data updates, the system ensures that it can adjust its recommendation strategies in real time based on market changes and user behavior.

[0049] To further improve system performance, the system employs caching optimization, query optimization, and indexing mechanisms to ensure fast recommendation response times under high concurrency. Simultaneously, the system also features adaptive learning capabilities, enabling dynamic adjustments and optimizations based on the effectiveness of recommendation results, such as introducing more refined dimensions of user profiles to enhance recommendation accuracy.

[0050] In summary, this invention, by introducing multi-strategy recall and online performance scoring, achieves accurate and intelligent content recommendation, bringing significant benefits in the following aspects: Enhancing User Experience and Platform Stickiness: Through multi-strategy recall integration, the system can accurately recommend content highly matched to user interests, thereby improving user experience and increasing platform stickiness. Personalized recommendations provide users with a personalized reading experience, reducing distractions from irrelevant information and improving the continuity and enjoyment of content consumption.

[0051] Boosting User Activity and Click-Through Rates: The online performance evaluation system dynamically adjusts recommendation weights based on real-time feedback on user click-through rates and interaction behavior, ensuring that recommended content aligns with users' current interests and trends, thereby significantly increasing user activity. Through precise content recommendations, user click-through rates and interaction frequency are significantly improved.

[0052] To achieve efficient distribution of diverse content and optimize content exposure, the system uses multi-strategy recall and differentiation coefficient calculation to balance the proportion of investor education, market trends, and product recommendations in the recommendation list. This effectively avoids information homogenization, achieves content diversity, and enables efficient distribution of various types of content. Especially for users with potential investment interests, the system prioritizes recommending market analysis, investor education, and information on popular products, enhancing content exposure.

[0053] Achieve precise user segmentation and differentiated operations: The system segments users based on dimensions such as content interests and dynamically adjusts recommended content for each segment based on online performance scores, thereby deeply satisfying users' personalized needs. For example, for users with high risk tolerance, the system recommends more high-yield products and market analysis; for conservative users, the recommended content leans towards risk warnings and investor education information, achieving differentiated operations.

[0054] Dynamically optimize the recommendation model to reduce system maintenance costs: The design that integrates real-time scoring and dynamic weights can adaptively optimize the recommendation model based on changes in user behavior, reducing reliance on manual parameter tuning and maintenance, and improving system stability and response speed. Example

[0055] This embodiment provides an intelligent recommendation system based on the fusion of online performance scoring and recall links, including a data collection module, a data preprocessing module, a multi-strategy recall module, an online performance scoring module, a weighted fusion and recommendation ranking module, a dynamic weight adjustment mechanism module, and an experimentation and verification module.

[0056] The data collection module is used to collect user historical behavior data, financial content information, and market trend data. User behavior data includes user click records, browsing history, and sharing history; financial content information includes content categories, tags, and publication time; market trend data covers financial news, industry dynamics, and policy changes.

[0057] The data preprocessing module cleans, deduplicates, and standardizes the collected data, handling missing and outlier values ​​to ensure data quality. Feature engineering is performed on user behavior data to extract user interest features, and text analysis is conducted on financial content data to extract keywords and topics, providing high-quality input for the recommendation model.

[0058] The multi-strategy recall module is used to implement multiple recall strategies based on user characteristics and historical behavior.

[0059] The online performance evaluation module calculates the online score for candidate products. The score is based on real-time feedback metrics such as click-through rate (CTR) and uniqueness rate (UR), using the following formula: ; Where i represents the content, j represents the specific content recall path, and UR represents the uniqueness score. Each candidate product calculates its corresponding online performance score based on different strategies.

[0060] Weighted Fusion and Recommendation Ranking Module: Weighted Average Score Calculation: The standardized scores from different content recall paths are weighted and summed to calculate the user's final recommendation score. Recommendation Ranking: Candidate content is ranked according to the final recommendation score, and the top N items with the highest scores are selected as the recommendation results.

[0061] Dynamic weight adjustment mechanism module: The system periodically adjusts the weight parameters of the recommendation algorithm based on user behavior feedback to improve recommendation accuracy.

[0062] Experiment and Validation Module: Online experiments are conducted in practical applications, randomly dividing users into two groups: the experimental group uses the recommendation system of this invention, and the control group uses a traditional recommendation system. The recommendation performance of the system is evaluated using various evaluation metrics (such as NDCG, MRR, MAP, precision, recall, F1 score, and AUC).

[0063] The specific functions of each module described above are explained in the relevant content of the method in Embodiment 1, and will not be repeated here. Example

[0064] This embodiment provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described in Embodiment 1. Example

[0065] This embodiment provides a computer device, including: Memory, used to store computer programs / instructions; A processor for executing the computer program / instructions to implement the steps of any of the methods described in Embodiment 1. Example

[0066] This embodiment provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of any of the methods described in Embodiment 1.

[0067] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

[0068] Those skilled in the art will understand that embodiments of this disclosure can be provided as methods, systems, or computer program products. Therefore, this disclosure can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this disclosure can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0069] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0070] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0071] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1The steps of the function specified in one or more boxes.

[0072] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this disclosure and not to limit its protection scope. Although this disclosure has been described in detail with reference to the above embodiments, those skilled in the art should understand that after reading this disclosure, they can still make various changes, modifications or equivalent substitutions to the specific implementation of the invention, but these changes, modifications or equivalent substitutions are all within the protection scope of the pending claims.

Claims

1. An intelligent recommendation method based on the fusion of online performance scoring and recall links, characterized in that, include: Set up multiple content recall links based on the content provided online and the content categories; Collect user data generated when users access content; The online performance score and weight parameters for each content recall link are calculated based on user data. The recommendation score for the content recall link is obtained by weighting and summing the online performance scores and weight parameters of the content recall link.

2. The intelligent recommendation method based on the fusion of online performance scoring and recall link as described in claim 1, characterized in that, The user data includes behavioral data generated by user access and click behavior within a preset time window, and content feature data obtained based on the attribute data of the accessed product.

3. The intelligent recommendation method based on the fusion of online performance scoring and recall link as described in claim 2, characterized in that, The preset time window is 60 days.

4. The intelligent recommendation method based on the fusion of online performance scoring and recall link as described in claim 2, characterized in that, The online performance score and weight parameters for each recall link are calculated based on user data, including: The user differentiation coefficient, which represents a user's preference for a certain type of content recall link, and the content differentiation coefficient, which represents the attention received by the content, are calculated based on user data. The online performance score and weight parameters for each recall link are calculated based on the user differentiation coefficient and the content differentiation coefficient.

5. The intelligent recommendation method based on the fusion of online performance scoring and recall link as described in claim 4, characterized in that, The user differentiation coefficient Content Differentiation Coefficient The mathematical expression is: ； ； In the formula, This represents the number of interactions between the i-th user and the k-th content retrieval link; This represents the total number of interactions between the i-th user and all content retrieval links; n is the total number of users, and m is the total number of content retrieval links; This represents the number of interactions of the k-th content retrieval link with the i-th user. This represents the total number of interactions among all users for the k-th content recall link.

6. The intelligent recommendation method based on the fusion of online performance scoring and recall link as described in claim 5, characterized in that, The online performance score and weight parameters for each content recall link are calculated based on the user differentiation coefficient and the content differentiation coefficient, including: The online performance of the content recall process is scored based on user differentiation coefficient, content differentiation coefficient, and content click-through rate. Standardize the online performance scores to obtain standardized output; The weight parameters for each content recall link are calculated based on the online performance score.

7. The intelligent recommendation method based on the fusion of online performance scoring and recall link as described in claim 6, characterized in that, The online performance evaluation of the content recall process is based on user differentiation coefficient, content differentiation coefficient, and content click-through rate, expressed as follows: ； In the formula, score represents the online performance rating, and r is the ranking of different content recall links after ranking based on content click-through rate. The ranking is sorted from highest to lowest based on the click-through rate of user behavior data.

8. The intelligent recommendation method based on the fusion of online performance scoring and recall link as described in claim 7, characterized in that, The online performance score is standardized to obtain the standardized output `normalized_score`, expressed as follows: ； In the formula, min(score) represents the lowest score among all online performance scores, and max(score) represents the highest score among all online performance scores.

9. The intelligent recommendation method based on the fusion of online performance scoring and recall link as described in claim 8, characterized in that, The weight parameter (weight) for each content recall link is calculated based on the online performance score, and its expression is: ； The standardized scores from different content retrieval paths are weighted and summed according to the weight parameters to calculate the user's final recommendation score, weighted_avg_score, as expressed in the following expression: 。 10. An intelligent recommendation system based on the fusion of online performance scoring and recall links, characterized in that, include: The data collection module is used to collect users' historical behavior data, financial content information, and market trend data; The data preprocessing module is used to preprocess the collected data to obtain processed data; The multi-strategy recall module is used to implement multiple recall strategies based on the content tag categories set according to the processed data and multiple content recall links. The online performance scoring module is used to calculate the online performance score and weight parameters for each content recall link based on user data generated when users access content. The weighted fusion and recommendation ranking module is used to sum the online performance scores of different content recall links according to the weight parameters to calculate the recommendation score; The dynamic weight adjustment mechanism module is used to periodically adjust weight parameters based on user behavior feedback.

Images