Search

CN-121526749-B - Electronic commerce platform interest analysis type commodity recommendation method based on artificial intelligence

CN121526749BCN 121526749 BCN121526749 BCN 121526749BCN-121526749-B

Abstract

The invention relates to the technical field of artificial intelligence e-commerce and discloses an interest analysis type commodity recommendation method of an e-commerce platform based on artificial intelligence. The method comprises the steps of constructing a network topology structure describing interest dynamic change based on user multidimensional behavior records, and an impromptu evolution network. Injecting a preset virtual interaction event into the network, simulating the propagation and attenuation of the event to deduce the potential interest development track of the user, and calculating the future liveness probability distribution of each interest node. And combining the platform real-time commodity supply flow and the environment attribute label thereof, mapping the commodity to an interest network, evaluating the matching strength, and then carrying out weighted correction by using the liveness probability distribution to generate a candidate commodity list with time weight. According to the invention, through dynamic networking modeling and virtual event simulation deduction, deeper understanding and more prospective grasp of user interests are realized, and the accuracy, diversity and timeliness of commodity recommendation are improved.

Inventors

  • HUANG BIN

Assignees

  • 莆田学院

Dates

Publication Date
20260508
Application Date
20260115

Claims (7)

  1. 1. An electronic commerce platform interest analysis type commodity recommendation method based on artificial intelligence is characterized by comprising the following steps: Establishing an interest evolution network for describing the dynamic change of the interest of a user, wherein the interest evolution network is constructed by multi-dimensional behavior records generated by the user in the electronic commerce platform in a time period, and carrying out time sequence association and mode abstraction on the multi-dimensional behavior records to form a network topological structure comprising a plurality of interest nodes and associated edges; Injecting a plurality of preset virtual interaction events into the interest evolution network, and calculating the probability distribution of the activity of each interest node in the interest evolution network in the future according to the potential interest development track by simulating the propagation path and attenuation process of the virtual interaction events in the interest evolution network and pushing the potential interest development track corresponding to the user; collecting a current real-time commodity supply flow of the e-commerce platform, wherein the real-time commodity supply flow comprises a commodity set and an attached environment attribute label, mapping the environment attribute label to the interest evolution network, evaluating the matching strength of each commodity in the commodity set and the interest node, and carrying out weighted correction on the matching strength according to the liveness probability distribution to generate a candidate commodity list with time weight; the matching strength is subjected to weighted correction according to the liveness probability distribution, and a candidate commodity list with time weight is generated, and the method comprises the following steps: Extracting probability values corresponding to a future specific time period in the liveness probability distribution; Performing fusion operation on the probability value and the initial matching strength of the commodity and the interest node; Reordering the commodities in the commodity set according to the fusion operation result; Adding a time weight coefficient for indicating recommended age urgency to each commodity after sorting; integrating the sequencing result and the time weight coefficient to generate the candidate commodity list with the time weight; Further comprises: Constructing an interest state comparison unit, wherein the interest state comparison unit is used for parallelly calculating stability indexes of the interest evolution network under different simulation intervention strategies, the simulation intervention strategies are formed by adding or inhibiting specific commodity information flows into the network topology structure, and the simulation intervention strategies which enable the stability indexes to reach optimal balance are selected as intervention strategies to be executed; Generating a commodity information pushing instruction set comprising pushing time and pushing sequences according to the intervention strategy to be executed and the candidate commodity list with the time weight, and inputting the commodity information pushing instruction to a recommendation interface of the electronic commerce platform; the selecting the simulation intervention strategy for enabling the stability index to reach the optimal balance as the intervention strategy to be executed comprises the following steps: the stability index comprises a network structure cohesive force index and an interest exploration index; calculating numerical combinations of the network structure cohesive force index and the interest exploration index under different simulation intervention strategies; positioning coordinate points corresponding to the numerical combination in a two-dimensional coordinate system; Calculating Euclidean distance from each coordinate point to a preset ideal balance point; And selecting a simulation intervention strategy associated with the coordinate point with the minimum Euclidean distance as the intervention strategy to be executed.
  2. 2. The method for recommending commodities according to an interest analysis method of an e-commerce platform based on artificial intelligence according to claim 1, wherein the step of performing time-series association and pattern abstraction on the multidimensional behavior record to form a network topology including a plurality of interest nodes and associated edges comprises: Stripping independent behavior events from the multidimensional behavior record and sorting the behavior events according to time stamps; Calculating the frequency and conditional probability of co-occurrence of different behavioral events in adjacent time windows; Clustering behavior events meeting co-occurrence conditions into interest clusters; Abstract characterizing each interest cluster as an interest node; and establishing association edges with directions and weights between the interest nodes according to the time sequence and co-occurrence relation between the behavior events, so as to form the network topological structure.
  3. 3. The method for artificial intelligence based e-commerce platform interest analysis type commodity recommendation according to claim 2, wherein said estimating a potential interest development track corresponding to said user comprises: Taking a currently active interest node in the interest evolution network as a starting point; Calculating the probability of interest state transition along the direction of the associated edge; Simulating the diffusion process of interest states along different paths by combining the content attribute of the virtual interaction event; Recording the activated new interest nodes and the activation intensity in the diffusion process; connecting paths with the activation intensity exceeding a threshold value to form the potential interest development track.
  4. 4. The method for recommending commodities according to an interest analysis method of an e-commerce platform based on artificial intelligence of claim 3, wherein the generating a set of commodity information push instructions including push timing and push sequence includes: Analyzing the intervention strategy to be executed, and identifying an interest node set requiring reinforcement in the strategy; screening a commodity subset with highest association degree with the interest node set from the candidate commodity list with time weight; Determining a pushing time interval corresponding to each commodity according to the time weight coefficient of each commodity in the commodity subset; in a pushing time interval, calculating specific pushing time by combining the historical online activity rule of the user; Arranging the commodity subsets according to the time sequence of the pushing opportunity to form the pushing sequence; and encoding the pushing time and the pushing sequence into commodity information pushing instructions which can be identified by the e-commerce platform recommendation interface.
  5. 5. The method for artificial intelligence based e-commerce platform interest analysis type commodity recommendation according to claim 4, further comprising: After the commodity information pushing instruction is executed, monitoring a secondary behavior sequence generated by the user aiming at the pushed commodity information, feeding the secondary behavior sequence back to the interest evolution network as a feedback signal, triggering parameter adjustment of corresponding interest nodes and associated edges in the interest evolution network, and completing one complete recommendation and feedback iteration.
  6. 6. The method for artificial intelligence based e-commerce platform interest analysis type commodity recommendation according to claim 5, wherein said monitoring the secondary behavior sequence generated by the user for the pushed commodity information comprises: capturing exposure, clicking, browsing duration, collection and shopping cart joining and purchasing behaviors of a user on the pushed commodity information; Encoding the captured behavior into a structured sequence of behavior types and time stamps according to the time sequence in which the behavior occurs; Calculating time intervals and behavior transfer modes between adjacent behaviors in the secondary behavior sequence; and taking the time interval and the behavior transfer mode as core contents of the feedback signal.
  7. 7. The method for recommending e-commerce platform interest analysis type commodity based on artificial intelligence according to claim 6, wherein said feeding back the secondary behavior sequence as a feedback signal to the interest evolution network, triggering parameter adjustment of corresponding interest nodes and associated edges in the interest evolution network, comprises: mapping the secondary behavior sequence back to the network topology structure, and identifying target interest nodes and associated edges affected by the sequence; According to the behavior type, determining the adjustment direction and the adjustment amplitude of the liveness parameter of the target interest node; According to the behavior transfer mode, determining the adjustment direction and the adjustment amplitude of the weight of the associated edge; and updating the internal state parameters of the interest evolution network by applying the adjustment direction and the adjustment amplitude.

Description

Electronic commerce platform interest analysis type commodity recommendation method based on artificial intelligence Technical Field The invention relates to the technical field of artificial intelligence electronic commerce, in particular to an electronic commerce platform interest analysis type commodity recommendation method based on artificial intelligence. Background Existing merchandise recommendation systems generally rely on user static portraits or time series behavioral models. The static portrayal method represents the user interest as a series of discrete, fixed labels or feature vectors that ignore the inherent associations between points of interest and their dynamic evolution over time. Models based on time sequence behavior sequences, such as a cyclic neural network or an attention model, can capture the sequence of behavior occurrence, but are essentially modeling the behavior sequences, and it is difficult to explicitly characterize and mine complex interaction and evolution relations between multiple concurrent interests of users. A common drawback of these approaches is that modeling of the user interests is isolated or linear, and the inability to effectively model the dynamic networked evolution process of interests as an organic whole results in inadequate understanding of the user interest changes, and difficulty in capturing potential transfer paths of interests. In the aspect of interest prediction, the conventional technology mostly adopts direct extrapolation or statistical probability prediction based on historical behavior patterns. Essentially based on the generalization and fitting of existing explicit behavior data, belonging to "retrospective" predictions. The method has the defects that the method is passively dependent on historical data, lacks the capability of active deduction and prospective simulation, and cannot discover and quantify the potential interest development direction in advance when the user does not generate related behaviors. When the user interest is in the sprouting or turning stage, the historical data is sparse or the mode is unknown, the prediction accuracy and the foresight of the method can be remarkably reduced. The invention aims to solve the technical problems of how to model a dynamic networking structure of user interests more accurately and how to actively simulate and deduce potential interest tracks of users. Disclosure of Invention The invention aims to provide an interest analysis type commodity recommendation method of an e-commerce platform based on artificial intelligence so as to solve the problems in the background technology. In order to achieve the above object, the present invention provides an interest analysis type commodity recommendation method for an e-commerce platform based on artificial intelligence, the method comprising: Establishing an interest evolution network for describing the dynamic change of the interest of a user, wherein the interest evolution network is constructed by multi-dimensional behavior records generated by the user in the electronic commerce platform in a time period, and carrying out time sequence association and mode abstraction on the multi-dimensional behavior records to form a network topological structure comprising a plurality of interest nodes and associated edges; Injecting a plurality of preset virtual interaction events into the interest evolution network, and calculating the probability distribution of the activity of each interest node in the interest evolution network in the future according to the potential interest development track by simulating the propagation path and attenuation process of the virtual interaction events in the interest evolution network and pushing the potential interest development track corresponding to the user; Collecting a current real-time commodity supply flow of the e-commerce platform, wherein the real-time commodity supply flow comprises a commodity set and an attached environment attribute label, mapping the environment attribute label to the interest evolution network, evaluating the matching strength of each commodity in the commodity set and the interest node, and carrying out weighted correction on the matching strength according to the liveness probability distribution to generate a candidate commodity list with time weight. Preferably, the performing time sequence association and pattern abstraction on the multidimensional behavior record to form a network topology structure including a plurality of interest nodes and associated edges includes: Stripping independent behavior events from the multidimensional behavior record and sorting the behavior events according to time stamps; Calculating the frequency and conditional probability of co-occurrence of different behavioral events in adjacent time windows; Clustering behavior events meeting co-occurrence conditions into interest clusters; Abstract characterizing each interest cluster as an interest