Search

CN-119831690-B - Preferential commodity personalized recommendation method and system

CN119831690BCN 119831690 BCN119831690 BCN 119831690BCN-119831690-B

Abstract

The invention provides a preferential commodity personalized recommendation method and a preferential commodity personalized recommendation system, which relate to the technical field of data processing, wherein the method comprises the steps of obtaining historical commodity browsing data of a target user; the method comprises the steps of establishing a weighted undirected graph reflecting interaction relation between a target user and each browsed commodity, calculating correlation scores of the target user and each browsed commodity in the weighted undirected graph through a restarting random walk algorithm with dynamic restarting probability, calculating similarity between commodity characteristics of each browsed commodity through Euclidean distance, determining bridging scores in inverse proportion relation with the similarity, determining recommendation scores of each browsed commodity according to the correlation scores and the bridging scores of the target user and each browsed commodity, calculating commodity characteristic similarity between each preferential commodity and the browsed commodity with the highest recommendation score through Mahalal distance, and pushing the preferential commodity to the target user from small to large in sequence. And the diversity of the recommendation results is improved while ensuring that the recommended preferential commodities accord with the user preference.

Inventors

  • FAN BIN

Assignees

  • 无锡闻游天下信息科技有限公司

Dates

Publication Date
20260508
Application Date
20241220

Claims (6)

  1. 1. The personalized recommendation method for the preferential commodities is characterized by comprising the following steps of: S1, acquiring historical commodity browsing data of a target user, wherein the historical commodity browsing data comprises a plurality of browsing commodities, commodity characteristics of each browsing commodity, browsing duration of each browsing commodity and clicking times of each browsing commodity; S2, establishing a weighted undirected graph reflecting the interaction relation between the target user and each browsed commodity based on the historical commodity browsing data, wherein nodes of the weighted undirected graph comprise target user nodes and browsed commodity nodes, edges between the nodes in the weighted undirected graph are the connection between the target user and each browsed commodity, and the weight of each edge is the preference degree obtained based on the click times and browsing time of each browsed commodity by the target user; S3, calculating the relevance scores of the target users and all browsed commodities in the weighted undirected graph through a restarting random walk algorithm with dynamic restarting probability based on the preference degree; s4, calculating the similarity between commodity features of all the browsed commodities through Euclidean distance, and determining bridging scores in inverse proportion relation with the similarity, wherein the bridging scores reflect preference degree spans of the target users on different browsed commodities; S5, determining the recommendation score of each browsed commodity according to the correlation score of the target user and each browsed commodity and the bridging score; s6, calculating the commodity feature similarity between each preferential commodity and the browsed commodity with the highest recommendation score through the Mahalanobis distance; S7, pushing the preferential commodity to the target user according to the sequence from small to large of the mahalanobis distance; the calculation mode of the relevance score specifically comprises the following steps: Wherein, the Representing the relevance score of the target user to the qth viewed item, The probability of a restart is indicated as such, Representing a unit vector describing the location of the qth browsed item in the weighted undirected graph, A preference degree matrix in a normalized form is represented, T represents the current walking number, T 0 represents the expected walking number, beta represents the adjustment coefficient, e represents the base number of natural logarithms, and D represents a degree matrix, wherein diagonal elements of the degree matrix are degrees of nodes; Wherein, the S4 specifically includes: s401, calculating similarity between commodity characteristics of each browsed commodity: Wherein, the Representing the similarity between the jth viewed item and the kth viewed item viewed by the ith target user, And The article feature vectors of the j-th and k-th browsed articles are respectively represented, Representing the gaussian kernel bandwidth parameter, Representing the square of the calculated euclidean distance, exp representing the natural exponential function; S402, calculating a similarity mean value: Wherein, the Representing the total number of items browsed by the ith target user, Representing the total relevance score between the browsed items browsed by the ith target user, Representing the average similarity between browsing commodities browsed by the ith target user; S403, determining the bridging score based on the similarity mean: Wherein, the The bridging score between browsing commodities browsed by the ith target user is represented, namely, the preference span of the ith target user for different browsing commodities is reflected; the calculation mode of the commodity feature similarity is specifically as follows: Wherein, the Representing the feature vector of the v-th preferential commodity, The browsing commodity feature vector with the highest recommendation score is represented, the corner mark T represents transposition, Representing the inverse of the covariance matrix between each of the preferential merchandise features and the browsing merchandise features, Representation of And (3) with The mahalanobis distance between the browsing commodities is the commodity feature similarity between the v-th preferential commodity and the browsing commodity with the highest recommendation score; the calculation mode of the recommendation score specifically comprises the following steps: Wherein, the Representing the recommendation score for the qth browsed item, Representing the relevance score of the target user to the qth viewed item, The bridging score between the browsing commodities browsed by the ith target user is represented to reflect the preference span of the ith target user for different browsing commodities.
  2. 2. The method for personalized recommendation of preferential commodities according to claim 1, wherein the weighted undirected graph is specifically: wherein V represents a node set comprising a target user node and a commodity browsing node, E represents an edge set comprising browsing time lengths of the target user for various purchased commodities, A represents a preference degree matrix of the target user for various commodity browsing, Indicating the preference degree of the target user i for browsing the commodity j, Represents the browsing duration of the target user i on the browsed commodity j, L represents the maximum browsing duration of the target user i on all browsed commodities, Representing the number of clicks of the target user i on the browsed commodity j, C representing the maximum number of clicks of the target user i on all browsed commodities, And Respectively representing browsing duration weight and click number weight, And Entropy indicating browsing time period and number of clicks respectively, And Representing a browsing commodity set and a purchased commodity set, respectively, G represents a weighted undirected graph with respect to V, E and W.
  3. 3. The method of claim 1, wherein the commodity characteristics include commodity price, commodity production date, and commodity type.
  4. 4. The method for personalized recommendation of preferential commodities according to claim 1, further comprising, after said S7: And updating the historical commodity browsing data at intervals of preset time.
  5. 5. A preferential commodity personalized recommendation system, comprising: A processor; A memory having stored thereon computer readable instructions which, when executed by the processor, implement the method of personalized recommendation of preferential merchandise of any one of claims 1 to 4.
  6. 6. A computer-readable storage medium, on which a computer program is stored, characterized in that the program, when executed by a processor, implements the method for personalized recommendation of preferential commodities according to any one of claims 1 to 4.

Description

Preferential commodity personalized recommendation method and system Technical Field The invention relates to the technical field of data processing, in particular to a preferential commodity personalized recommendation method and system. Background Commodity recommendation is a technique that utilizes algorithms and user data to recommend commodities to users that they may be interested in. The method helps the user to find potential interesting commodities by analyzing browsing, clicking and purchasing behaviors of the user and combining the properties of the commodities and behaviors of other users, and improves shopping experience. The commodity recommendation plays a key role in connecting a user and commodities in mass commodity information, and the problem that the user is difficult to select in an information overload environment is solved. The method not only can improve user experience and help users to quickly find required commodities, but also can increase sales of the platform, and attracts users to purchase through personalized recommendation so as to optimize the income structure of the platform. However, the existing commodity recommendation method is excessively limited to the historical purchasing records of the users, so that the over-recommended and historical purchasing commodities and the commodities close to the historical purchasing commodities are excessively concentrated, and the lack of diversity may cause the users to lose interest in repeated recommendation, so that the viscosity of the platform is reduced. Disclosure of Invention The invention provides a preferential commodity personalized recommendation method and a preferential commodity personalized recommendation system, which aim to solve the technical problems that in the prior art, a commodity recommendation method is excessively limited to a historical purchasing record of a user, so that excessive recommendation, historical purchasing of commodities and commodities close to the historical purchasing of the commodities are excessively concentrated, and repeated recommendation is possibly lost due to lack of diversity. The technical scheme provided by the embodiment of the invention is as follows: First aspect The personalized recommendation method for preferential commodities provided by the embodiment of the invention comprises the following steps: s1, acquiring historical commodity browsing data of a target user, wherein the historical commodity browsing data comprises a plurality of browsing commodities, commodity characteristics of each browsing commodity, browsing duration of each browsing commodity and clicking times of each browsing commodity; s2, establishing a weighted undirected graph reflecting interaction relation between a target user and each browsed commodity based on historical commodity browsing data, wherein nodes of the weighted undirected graph comprise target user nodes and browsed commodity nodes, edges between the nodes in the weighted undirected graph are the connection between the target user and each browsed commodity, and the weight of each edge is the preference degree obtained based on the click times and browsing time of each browsed commodity by the target user; s3, calculating the relevance scores of the target users and all browsed commodities in the weighted undirected graph through a restarting random walk algorithm with dynamic restarting probability based on the preference degree; S4, calculating correlation among commodity features of all the browsed commodities through Euclidean distance, and determining bridging scores in inverse proportion relation with the correlation, wherein the bridging scores reflect preference degree spans of target users on different browsed commodities; S5, determining the recommendation score of each browsed commodity according to the correlation score and the bridging score of the target user and each browsed commodity; s6, calculating the commodity feature similarity between each preferential commodity and the browsed commodity with the highest recommendation score through the Mahalanobis distance; and S7, pushing the preferential commodity to the target user according to the sequence from big to small in the mahalanobis distance. Second aspect The personalized recommendation system for preferential commodities provided by the embodiment of the invention comprises the following components: A processor; And a memory having stored thereon computer readable instructions which, when executed by the processor, implement the method for personalized recommendation of preferential merchandise according to the first aspect. Third aspect of the invention An embodiment of the present invention provides a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the method for personalized recommendation of preferential merchandise according to the first aspect. The technical scheme provided by the e