User dynamic classification-based e-commerce platform commodity recommendation method and system

Abstract

The invention discloses a user dynamic classification-based e-commerce platform commodity recommendation method and system. The method comprises the steps of obtaining basic information, real-time comment data and real-time browsing records of to-be-recommended users; extracting an interest feature tag vector and a vector weight of the to-be-recommended users from the acquired data; classifying the to-be-recommended users based on the interest feature tag vectors and the vector weights of the to-be-recommended users to obtain user categories of the to-be-recommended users; and outputting recommended commodities based on the user category of the to-be-recommended users and a pre-constructed corresponding relationship list of the user category and the commodity category.

Description

technical field [0001] The present disclosure relates to the technical field of commodity recommendation, in particular to a method and system for recommending commodities on an e-commerce platform based on user dynamic classification. Background technique [0002] The statements in this section merely mention background art related to the present disclosure and do not necessarily constitute prior art. [0003] On the e-commerce platform, there are a wide variety of commodities, and user behaviors include searching, browsing, clicking, collecting, purchasing, etc., which constitute user behavior logs. The log data of different users and items vary widely, and the data volume of click behavior is often much greater than that of purchase behavior. The amount of data, therefore, only based on purchase and collection behaviors cannot accurately perceive the purchase intention of online customers. Using the multi-dimensional data of product sales, customer registration, and purch...

AI Technical Summary

Problems solved by technology

[0017] 1. The decision factors for making recommendations to users are relatively single, mainly based on the user's historical purchase and browsing behavior, without comprehensive consideration of the overall historical behavior of the user's category;

[0018] 2. The product recommendation list is generated based on fixed rules, which do not have the ability of dynamic self-adjustment, and the recommendation accuracy will drop significantly after long-term operation;

[0019] 3. In user classification, research is often conducted only from the user’s static preferences, ignoring changes in user interests, and the weight of user interest tags cannot be updated in time, resulting in a large degree of ambiguity in user classification, causing recommendation behavior to consider user types and this type The user's preference changes insufficiently, resulting in the decline of recommendation accuracy and missing recommendation problems;

[0020] 4. The calculation of matching similarity between products and users does not take into account influencing factors such as user type, product type, attention time, price fluctuations, promotional activities, etc., resulting in inaccurate calculation conclusions, which makes users feel "garbage recommendation"

Images