Personalized commodity recommendation method and system based on collaborative filtering algorithm

Abstract

The invention relates to the technical field of commodity recommendation, in particular to a personalized commodity recommendation method and system based on a collaborative filtering algorithm. Analyzing and processing the user data; constructing a user and commodity scoring matrix; calculating the similarity between the users, and generating a user set with high similarity; generating a recommended commodity table through the user set with high similarity; displaying the recommended commodity to the new user; user feedback is created to facilitate subsequent optimization; according to the method, positive feedback commodity recommendation in the similar user set is recommended to the new user, and similarly, new products are also recommended to the user set with high similarity, so that the use experience of the new user is improved, and the popularization of the new products is promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of commodity recommendation, and in particular to a personalized commodity recommendation method and system based on a collaborative filtering algorithm. Background Art

[0002] With the rapid development of e-commerce, personalized product recommendation systems have become an important means for e-commerce platforms to improve user experience and increase sales.

[0003] In the prior art, existing personalized product recommendation systems mostly use collaborative filtering algorithms, which can analyze users' historical behavior data (such as purchase records, browsing records, etc.) and recommend products that users may be interested in. However, existing personalized product recommendation systems based on collaborative filtering algorithms still have some problems and shortcomings. For example, new users or new products lack sufficient historical data to support the calculation of the recommendation algorithm, resulting in the recommendation algorithm being unable to accurately capture the characteristics and interests of new users or new products. Summary of the invention

[0004] The present invention provides a personalized product recommendation method and system based on a collaborative filtering algorithm. By constructing a user and product rating matrix and a similarity calculation module, a user set with similar interests is matched for a new user or a new product. Positive feedback products in the similar user set are recommended to the new user. Similarly, new products are also recommended to the user set with high similarity, thereby improving the user experience of the new user and promoting the promotion of new products.

[0005] To achieve the above objectives, the present invention provides a personalized product recommendation method and system based on a collaborative filtering algorithm. The technical solutions adopted are as follows:

[0006] In a first aspect, an embodiment of the present invention provides a personalized product recommendation method based on a collaborative filtering algorithm, the method comprising:

[0007] S1. Collection of user data: Collect user data from different sources, including product clicks, product browsing, product purchases, and product ratings;

[0008] S2. Process and analyze user data: Use the distributed computing framework Hadoop to process and analyze user data;

[0009] S3. Constructing a user and product rating matrix for the processed and analyzed user data: After determining the matrix elements, fill the matrix elements with the processed and analyzed user data to complete the construction of the user and product rating matrix;

[0010] S4. Calculate the similarity between users and find a set of users with similar interests to the new user: by calculating the similarity between different user behavior data or the similarity between items, and sorting the calculated user similarities, a set of users with similar interests to the new user is constructed;

[0011] S5. Generate a list of recommended products based on the browsing behaviors of similar users in a user set with similar interests to the new user: match the user set with similar interests to the user and product rating matrix, prioritize them, and finally generate a list of recommended products;

[0012] S6. Display recommended products to new users: Create a new recommendation style to facilitate users to select recommended products;

[0013] S7. Create user feedback to facilitate subsequent optimization: After a new user selects a product, user feedback is provided and corresponding optimization is performed based on the feedback.

[0014] Furthermore, the collection of the user data specifically includes the following steps:

[0015] S21. Filter user data: filter the collected user data to remove unnecessary data;

[0016] S22. User data conversion format: Convert the filtered user data into a format that can be processed by Hadoop.

[0017] Furthermore, the user data is processed and analyzed, specifically including the following steps:

[0018] S31. Establishing a processing model: Installing Hadoop and configuring the cluster, uploading the filtered and formatted user data to the Hadoop distributed file system;

[0019] S32. Design a data processing program and run it: Run the program on the Hadoop cluster. Hadoop will distribute the data to different nodes for processing. After the program is finished running, the captured data will be sent to the back-end server through the API or data pipeline for storage and analysis to obtain the data between users and products.

[0020] Furthermore, constructing the user and product rating matrix specifically includes the following steps:

[0021] S41. Determine the elements of the user and product rating matrix: user set, user set is represented by U = u 1 ,u 2 ,u 3 ,…,u m Represents, where m is the number of users;

[0022] Commodity set, commodity set is represented by I=i 1 ,i 2 ,i 3 ,…,i n Represents, where n is the number of items;

[0023] Rating value: The rating value indicates the user's preference or evaluation of the product, expressed in numbers;

[0024] S42. Construct user and product rating matrix: Create an m×n two-dimensional matrix, where rows represent users and columns represent products. The elements R in the matrix are ij represents the rating of the i-th user on the j-th product;

[0025] S43. Fill the processed user program into the user and product rating matrix: fill the elements in the matrix according to the historical behavior data of the user data.

[0026] Furthermore, the personalized product recommendation method based on the collaborative filtering algorithm also includes optimizing and adjusting the user and product rating matrix, specifically including the following steps:

[0027] S51. Dimension reduction of the user and product rating matrix: dimensionality reduction processing of the user and product rating matrix is ​​performed by using dimensionality reduction techniques (such as matrix decomposition, singular value decomposition, etc.);

[0028] S52. Filling the missing values ​​in the user and product rating matrix: filling the missing values ​​in the user and product rating matrix by filling based on user similarity, filling based on item similarity, and filling based on matrix dimensionality reduction.

[0029] Furthermore, the similarity between users is calculated to find a set of users with similar interests to the new user, which specifically includes the following steps:

[0030] S61. Similarity calculation: Based on user behavior data (user browsing behavior, etc.), the cosine similarity calculation method is used to calculate the similarity between users or the similarity between items. The specific formula is:

[0031]

[0032] Where A and B are the vectors of two users, θ is the angle between the two users, (A·B) is the dot product of the two users, |A| and |B| are the moduli of the two users respectively;

[0033] S62. Sorting similarity: sorting the calculated user similarities and finding the top K users with the highest similarity to the new user (K ​​is a preset threshold);

[0034] S63. Construct a similar user set: group the first K sorted users into a set.

[0035] Furthermore, generating a list of recommended products based on the browsing behaviors of similar users in a user set with similar interests to the new user specifically includes the following steps:

[0036] S71. Determine the recommended product list: Substitute the user set with high similarity to the new user into the user and product rating matrix, and predict the product list with high ratings among the user set with high similarity to the new user;

[0037] S72. Pre-select priority recommended products: pre-select the top K users (i.e. similar users) who are most similar to the new user, and recommend a product to the new user based on the evaluations of these similar users on the product, or directly recommend products that the similar users like and that the new user has not yet known to the new user;

[0038] S73. Prioritize the recommended product list: After calculating the similarity between product items, obtain the list of positive feedback items in the user's historical behavior data (i.e., products that the user likes), and use the collaborative filtering algorithm to find the top K similar products (similar product set), then sort the similar product set according to the similarity score, and generate the final recommendation list.

[0039] Furthermore, displaying a list of recommended products to the new user specifically includes the following steps:

[0040] S81. Data processing of the recommended product list: convert the recommended product list into recommendation data, and filter it through collaborative filtering algorithm to remove duplicate, invalid or missing key information products, and then extract and organize the product click, description, rating and picture link information according to display needs;

[0041] S82. Display recommended product information to users: different product display methods, product information arrangement order and product presentation methods are convenient for users to understand.

[0042] Furthermore, the personalized product recommendation method based on the collaborative filtering algorithm also includes displaying the recommended product list to the new user for incremental training, which specifically includes the following steps:

[0043] S91. Load the trained recommendation model from the data storage module, use the newly collected data as incremental samples, and perform incremental training on the recommendation model;

[0044] S92. After incremental training, the recommendation model is evaluated using the validation set to ensure that the performance of the recommendation model has not significantly decreased;

[0045] S93. Fine-tune the recommendation model according to the evaluation results, and store the updated recommendation model parameters back to the storage medium for subsequent use.

[0046] In a second aspect, another embodiment of the present invention provides a personalized product recommendation system based on a collaborative filtering algorithm, the system comprising:

[0047] A data collection module, used to collect multiple different types of user data, including platform type and user type;

[0048] A distributed computing module is used to install Hadoop and configure a cluster to speed up the response time for processing the user data;

[0049] A similarity calculation module is used to calculate the similarity between multiple users or the similarity between commodity items;

[0050] The recommendation generation module generates a list of recommended products based on the similarity calculation results and the browsing behaviors of similar users in the user set with similar interests to the new user;

[0051] Collaborative filtering module, which selects a set of similar items with high similarity from the list of positive feedback items;

[0052] Data storage module, used to load recommendation models and data storage;

[0053] The recommendation result display module includes a display unit, a data integration unit and a user interaction unit. The display unit provides users with different display methods of commodities, arrangement order and presentation methods of information to facilitate user understanding. The data integration unit is connected with the data storage module to realize dynamic loading and updating of data and improve the response speed of the display unit. The user interaction unit gives the display unit corresponding interaction logic and connects to the data integration unit.

[0054] Compared with the prior art, in the embodiment of the present invention, the distributed computing framework Hadoop is adopted to analyze and process the user data, and at the same time, a user and product rating matrix is ​​established. After the processed user data is substituted into the user and product rating matrix, the user's rating value for the product is obtained, which provides data support for subsequent product recommendations. The similarity technology is introduced again, and the user set with high similarity to the new user is calculated through similarity. The user set data with high similarity is sent to the recommendation model. The recommendation model generates a recommended product table according to the user and product rating matrix. The recommendation model then screens out the products with positive feedback in the recommended product table, and finally transmits them to the display module for recommendation to the new user, thereby providing accurate product recommendations to the new user, improving the new user experience, and enhancing the new user stickiness. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 A flowchart of a personalized product recommendation method based on a collaborative filtering algorithm provided by an embodiment of the present invention;

[0056] Figure 2 A structural diagram of a personalized product recommendation system based on a collaborative filtering algorithm provided by one embodiment of the present invention;

[0057] Figure 3 A schematic diagram of a personalized product recommendation method and system based on a collaborative filtering algorithm provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0058] In order to provide a further understanding of the purpose, structure, features, and functions of the present invention, the following detailed description is given in conjunction with the embodiments.

[0059] The following is a detailed description of a personalized product recommendation method and system based on a collaborative filtering algorithm provided by the present invention with reference to the accompanying drawings.

[0060] Embodiment 1:

[0061] The present invention provides a personalized product recommendation method based on collaborative filtering algorithm, such as Figure 1 As shown, the following steps are included:

[0062] S1. Collection of user data: Collect user data from different sources, including product clicks, product browsing, product purchases, and product ratings;

[0063] Collect user data from different sources, including product clicks, product browsing, product purchases, and product ratings. Product clicks, product browsing, and product purchases are all used to determine the popularity of products. Product ratings are used to determine the popularity and quality of products, providing data for subsequent product recommendations.

[0064] After completing the above data collection, perform data filtering and cleaning to remove unnecessary data, and convert the data format to a format that Hadoop can process, such as XML or JSON format.

[0065] S2. Process and analyze user data: Use the distributed computing framework Hadoop to process and analyze user data; Use the distributed computing framework Hadoop and configure a cluster to process user data, so as to achieve rapid processing of large amounts of data and shorten the user data processing time;

[0066] By setting up a program to process data using the MapReduce programming model on the Hadoop cluster, and running the program on the Hadoop cluster when processing data, Hadoop will distribute user data to different nodes for processing. After the program is finished running, the captured user data will be sent to the back-end server through the API or data pipeline for storage and analysis.

[0067] S3. Constructing a user and product rating matrix for the processed and analyzed user data: After determining the matrix elements, fill the matrix elements with the processed and analyzed user data to complete the construction of the user and product rating matrix;

[0068] Determine the three elements in the user and product rating matrix, the user set, and the user set is represented by U = u 1 ,u 2 ,u 3 ,…,u m Indicated by, where m is the number of users; the commodity set, the commodity set is represented by I = i 1 ,i 2 ,i 3 ,…,i n , where n is the number of items; the rating value, which indicates the user's preference or evaluation of the product, is expressed as a number, such as an integer between 1 and 5. The larger the number, the more the user likes the product or the higher the evaluation;

[0069] Thus, an m×n two-dimensional matrix is ​​created, where the rows represent the above users and the columns represent the above products. The elements R in the matrix are ij represents the above rating value of the i-th user for the j-th product;

[0070] After the construction is completed, the captured user data will be sent by the back-end server (data storage device) to fill in the rating values ​​in the user and product rating matrix. When the user does not have a clear rating value for the product, but has other behavior data (such as browsing, purchasing, etc.), the rating value can be estimated based on these behavior data or processed in other ways (such as setting it as a default value or ignoring it by the system).

[0071] Furthermore, the personalized product recommendation method based on the collaborative filtering algorithm also includes optimizing and adjusting the user and product rating matrix, which specifically includes the following steps:

[0072] In practical applications, the user-product rating matrix may be sparse, that is, most of the elements are null values. In order to optimize the recommendation effect, it is necessary to reduce the dimension of the user and product rating matrix: use dimensionality reduction techniques (such as matrix decomposition, singular value decomposition, etc.) to reduce the dimension of the user and product rating matrix;

[0073] Filling the missing values ​​in the user and product rating matrix again: filling the missing values ​​in the user and product rating matrix by filling based on user similarity, filling based on item similarity, and filling based on matrix dimensionality reduction;

[0074] Finally, by comparing the recommendation results with the user's actual behavior data, we can verify the effect of the recommended product list in user feedback, and optimize and adjust the matrix construction method based on the verification results.

[0075] S4. Calculate the similarity between users and find a set of users with similar interests to the new user: by calculating the similarity between different user behavior data or the similarity between items, and sorting the calculated user similarities, a set of users with similar interests to the new user is constructed;

[0076] Based on user behavior data (user browsing behavior, etc.), the cosine similarity calculation method is used to calculate the similarity between users or the similarity between items. The specific formula is:

[0077]

[0078] Where A and B are the vectors of two users, θ is the angle between the two users, (A·B) is the dot product of the two users, |A| and |B| are the moduli of the two users respectively;

[0079] The calculated user similarities are sorted to find the top K users (K is a preset threshold) with the highest similarity to the new user, and the sorted top K users are grouped into a set to form a data reference.

[0080] S5. Generate a list of recommended products based on the browsing behaviors of similar users in a user set with similar interests to the new user: match the user set with similar interests to the user and product rating matrix, prioritize them, and finally generate a list of recommended products;

[0081] Determine the recommended product list: Substitute the user set with high similarity to the new user into the user and product rating matrix, and predict the product list with high ratings from the user set with high similarity to the new user;

[0082] Pre-select and prioritize recommended products: Pre-select the top K users (i.e. similar users) who are most similar to the new user, and recommend them to the new user based on their evaluation of a product. Or directly recommend products that similar users like and that the new user has not yet learned about to the new user.

[0083] Prioritize the recommended product list: After calculating the similarity between product items, obtain the list of positive and negative feedback items in the user's historical behavior data (i.e., products that the user likes), and use the collaborative filtering algorithm to find the top K similar products (a set of similar products). Then, sort the set of similar products according to the similarity score and generate the final recommendation list.

[0084] S6. Display recommended products to new users: Create a new recommendation style to facilitate users to select recommended products;

[0085] Data processing of recommended product lists: convert the recommended product lists into recommendation data, and filter them through collaborative filtering algorithms to remove duplicate, invalid or missing key information products, and then extract and organize information such as product clicks, descriptions, ratings and image links according to display needs;

[0086] Display recommended product information to users: different product display methods, product information arrangement order and product presentation methods are easy for users to understand;

[0087] Furthermore, the personalized product recommendation method based on the collaborative filtering algorithm also includes displaying the recommended product list to the new user for incremental training, which specifically includes the following steps:

[0088] Load the trained recommendation model from the data storage module, use the newly collected user data as incremental samples, and perform incremental training on the recommendation model, which usually involves using stochastic gradient descent (SGD) or its variants to perform gradient updates on the new samples to adjust the model parameters;

[0089] After incremental training, the recommendation model is evaluated using the validation set to ensure that the performance of the recommendation model has not significantly decreased;

[0090] At the same time, the recommendation model is fine-tuned according to the evaluation results, and the updated recommendation model parameters are stored back to the storage medium for subsequent use.

[0091] The principle of embodiment 1 of the present invention is, for example, Figure 3 As shown in the figure, user C (new user) has browsed item A. At this time, it is collected that both user A and user B have browsed item A, so user A and user B are determined to be a user set with high similarity. At this time, the items browsed by user A and user B are provided to user C. At this time, the collaborative filtering algorithm is used to obtain the final recommended product table, and the recommended product is item C. Therefore, item C is recommended to user C (new user).

[0092] It should be noted that new products and new users are of the same type. When a user browses, purchases and evaluates a new product, the collaborative filtering algorithm in Example 1 of the present invention evaluates the new product to determine whether the new product is a positive feedback product. When the new product is a positive feedback product, the new product will be recommended to a set of users with high similarity to the users who have purchased and browsed the product, thereby increasing the promotion of the new product.

[0093] Embodiment 2:

[0094] The present invention proposes a personalized product recommendation system based on collaborative filtering algorithm. Figure 2 As shown, it shows a structure diagram of a personalized product recommendation system based on a collaborative filtering algorithm provided by an embodiment of the present invention, the system comprising:

[0095] A data collection module, used to collect multiple different types of user data, including platform type and user type;

[0096] A distributed computing module is used to install Hadoop and configure a cluster to speed up the response time for processing the user data;

[0097] A similarity calculation module is used to calculate the similarity between multiple users or the similarity between commodity items;

[0098] The recommendation generation module generates a list of recommended products based on the similarity calculation results and the browsing behaviors of similar users in the user set with similar interests to the new user;

[0099] Collaborative filtering module, which selects a set of similar items with high similarity from the list of positive feedback items;

[0100] Data storage module, used to load recommendation models and data storage;

[0101] The recommendation result display module includes a display unit, a data integration unit and a user interaction unit. The display unit provides users with different display methods of commodities, arrangement order and presentation methods of information to facilitate user understanding. The data integration unit is connected with the data storage module to realize dynamic loading and updating of data and improve the response speed of the display unit. The user interaction unit gives the display unit corresponding interaction logic and connects to the data integration unit.

[0102] Preferably, by setting up a data storage module, when the data is cached by querying the cache using the data storage module, when the user initiates a data query request, first check whether the required data is in the cache. If the data exists in the cache and is not expired, the cached data is directly returned to reduce the access pressure of the data storage module or remote service. If the data does not exist in the cache or the data is expired, the data is obtained from the data storage module or remote service, and the obtained data is stored in the cache again, and a corresponding expiration time is set. When the accessed data is updated, the old data in the cache needs to be deleted at the same time to ensure that the user can obtain the latest data when querying again, thereby caching the popular recommendation results in the data storage module, reducing the computational overhead of each recommendation, further improving the real-time performance of the recommendation system, and reducing the load on the server.

[0103] Preferably, a cache invalidation strategy (such as LRU, LFU, etc.) is used to automatically manage cache space to avoid cache data expiration or invalid occupation.

[0104] It should be noted that the device provided in the above embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device is divided into different functional modules to complete all or part of the functions described above. In addition, the system for personalized product recommendation based on collaborative filtering algorithm and the method embodiment for personalized product recommendation based on collaborative filtering algorithm provided in the above embodiment belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0105] In the description of the present invention, it is necessary to understand that the terms “longitudinal”, “lateral”, “length”, “width”, “thickness”, “up”, “down”, “front”, “back”, “left”, “right”, “vertical”, “horizontal”, “top”, “bottom”, “inside” and “outside” orientations or positional relationships are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention.

[0106] The present invention has been described by the above-mentioned relevant embodiments, however, the above-mentioned embodiments are only examples for implementing the present invention. It must be pointed out that the disclosed embodiments do not limit the scope of the present invention. On the contrary, changes and modifications made without departing from the spirit and scope of the present invention are all within the scope of patent protection of the present invention.

Claims

1. A personalized product recommendation method based on collaborative filtering algorithm, characterized in that: The personalized product recommendation method based on collaborative filtering algorithm comprises the following steps: S1. Collection of user data: Collect user data from different sources, including product clicks, product browsing, product purchases, and product ratings; S2. Process and analyze user data: Use the distributed computing framework Hadoop to process and analyze user data; S3. Constructing a user and product rating matrix for the processed and analyzed user data: After determining the matrix elements, fill the matrix elements with the processed and analyzed user data to complete the construction of the user and product rating matrix; S4. Calculate the similarity between users and find a set of users with similar interests to the new user: by calculating the similarity between different user behavior data or the similarity between items, and sorting the calculated user similarities, a set of users with similar interests to the new user is constructed; S5. Generate a list of recommended products based on the browsing behaviors of similar users in a user set with similar interests to the new user: match the user set with similar interests to the user and product rating matrix, prioritize them, and finally generate a list of recommended products; S6. Display recommended products to new users: Create a new recommendation style to facilitate users to select recommended products; S7. Create user feedback to facilitate subsequent optimization: After a new user selects a product, user feedback is provided and corresponding optimization is performed based on the feedback.

2. The personalized product recommendation method based on collaborative filtering algorithm according to claim 1, characterized in that: The collection of the user data specifically includes the following steps: S21. Filter user data: filter the collected user data to remove unnecessary data; S22. User data conversion format: Convert the filtered user data into a format that can be processed by Hadoop.

3. The personalized product recommendation method based on collaborative filtering algorithm according to claim 1, characterized in that: Processing and analyzing the user data specifically includes the following steps: S31. Establishing a processing model: Installing Hadoop and configuring the cluster, uploading the filtered and formatted user data to the Hadoop distributed file system; S32. Design a data processing program and run it: Run the program on the Hadoop cluster. Hadoop will distribute the data to different nodes for processing. After the program is finished running, the captured data will be sent to the back-end server through the API or data pipeline for storage and analysis to obtain the data between users and products.

4. The personalized product recommendation method based on collaborative filtering algorithm according to claim 3 is characterized in that: The user and product rating matrix is ​​constructed, specifically including the following steps: S41. Determine the elements of the user and product rating matrix: user set, user set is represented by U = u1, u2, u3, ..., u m Represents, where m is the number of users; Commodity set, commodity set is I = i1, i2, i3, ..., i n Represents, where n is the number of items; Rating value: The rating value indicates the user's preference or evaluation of the product, expressed in numbers; S42. Construct user and product rating matrix: Create an m×n two-dimensional matrix, where rows represent users and columns represent products. The elements R in the matrix are ij represents the rating of the i-th user on the j-th product; S43. Fill the processed user program into the user and product rating matrix: fill the elements in the matrix according to the historical behavior data of the user data.

5. The personalized product recommendation method based on collaborative filtering algorithm according to claim 4 is characterized in that: The method also includes optimizing and adjusting the user and product rating matrix, which specifically includes the following steps: S51. Dimension reduction of the user and product rating matrix: dimensionality reduction processing of the user and product rating matrix is ​​performed by using dimensionality reduction techniques (such as matrix decomposition, singular value decomposition, etc.); S52. Filling the missing values ​​in the user and product rating matrix: filling the missing values ​​in the user and product rating matrix by filling based on user similarity, filling based on item similarity, and filling based on matrix dimensionality reduction.

6. The personalized product recommendation method based on collaborative filtering algorithm according to claim 1, characterized in that: Calculate the similarity between users and find a set of users with similar interests to the new user. Specifically, the following steps are included: S61. Similarity calculation: Based on user behavior data (user browsing behavior, etc.), the cosine similarity calculation method is used to calculate the similarity between users or the similarity between items. The specific formula is: Where A and B are the vectors of two users, θ is the angle between the two users, (A·B) is the dot product of the two users, |A| and |B| are the moduli of the two users respectively; S62. Sorting similarity: sorting the calculated user similarities and finding the top K users with the highest similarity to the new user (K ​​is a preset threshold); S63. Construct a similar user set: group the first K sorted users into a set.

7. The personalized product recommendation method based on collaborative filtering algorithm according to claim 1, characterized in that: Generate a list of recommended products based on the browsing behaviors of similar users in a user set with similar interests to the new user, specifically including the following steps: S71. Determine the recommended product list: Substitute the user set with high similarity to the new user into the user and product rating matrix, and predict the product list with high ratings among the user set with high similarity to the new user; S72. Pre-select priority recommended products: pre-select the top K users (i.e. similar users) who are most similar to the new user, and recommend a product to the new user based on the evaluations of these similar users on the product, or directly recommend products that the similar users like and that the new user has not yet known to the new user; S73. Prioritize the recommended product list: After calculating the similarity between product items, obtain the list of positive and negative feedback items in the user's historical behavior data (i.e., products that the user likes), and use the collaborative filtering algorithm to find similar TopK products (a set of similar products). Then, sort the set of similar products according to the similarity score and generate the final recommendation list.

8. The personalized product recommendation method based on collaborative filtering algorithm according to claim 1, characterized in that: Displaying a list of recommended products to the new user specifically includes the following steps: S81. Data processing of the recommended product list: convert the recommended product list into recommendation data, and filter it through collaborative filtering algorithm to remove duplicate, invalid or missing key information products, and then extract and organize the product click, description, rating and picture link information according to display needs; S82. Display recommended product information to users: different product display methods, product information arrangement order and product presentation methods are convenient for users to understand.

9. The personalized product recommendation method based on collaborative filtering algorithm according to claim 8, characterized in that: The method further includes displaying the recommended product list to the new user for incremental training, which specifically includes the following steps: S91. Load the trained recommendation model from the data storage module, use the newly collected data as incremental samples, and perform incremental training on the recommendation model; S92. After incremental training, the recommendation model is evaluated using the validation set to ensure that the performance of the recommendation model has not significantly decreased; S93. Fine-tune the recommendation model according to the evaluation results, and store the updated recommendation model parameters back to the storage medium for subsequent use.

10. A personalized product recommendation system based on collaborative filtering algorithm, characterized in that: The system includes: A data collection module, used to collect multiple different types of user data, including platform type and user type; A distributed computing module, used to install Hadoop and configure a cluster to speed up the response time for processing the user data; A similarity calculation module is used to calculate the similarity between multiple users or the similarity between commodity items; The recommendation generation module generates a list of recommended products based on the similarity calculation results and the browsing behaviors of similar users in the user set with similar interests to the new user; Collaborative filtering module, which selects a set of similar items with high similarity from the list of positive feedback items; Data storage module, used to load recommendation models and data storage; The recommendation result display module includes a display unit, a data integration unit and a user interaction unit. The display unit provides users with different display methods of commodities, arrangement order and presentation methods of information to facilitate user understanding. The data integration unit is connected with the data storage module to realize dynamic loading and updating of data and improve the response speed of the display unit. The user interaction unit gives the display unit corresponding interaction logic and connects to the data integration unit.

Images