Multi-e-commerce cross recommendation method based on clustering feature migration

Abstract

The invention discloses a multi-e-commerce cross recommendation method based on clustering feature migration. The method comprises the following steps: 1) score matrix construction stage: a, collecting e-commerce data; b, data cleaning and de-noising; c, constructing a score matrix, and d end; 2) auxiliary domain learning stage: a, acquiring the score matrix; b, extracting a user / project feature matrix; c, clustering the user / project feature matrix; d, computing average score; e, constructing a clustering feature matrix; and f, repeating the above steps till to end for each auxiliary e-commerce; and 3) target domain learning stage: a, acquiring a target e-commerce score matrix; b, migrating the clustering feature to accomplish matrix decomposition; c, reconstructing a target e-commerce score matrix; d, producing a recommendation list; and e, ending. And new solution thought is provided for solving the data sparsity, cold-starting and dilemmatic diversity and precision problem existingin the e-commerce recommendation system by using the migration learning technology.

Description

technical field [0001] The invention relates to a multi-e-commerce cross-recommendation method, which solves the problem that the e-commerce recommendation system has low recommendation accuracy in the case of extremely sparse data and cold start. Background technique [0002] With the continuous expansion of e-commerce sites, the problem of information overload is becoming more and more serious. A very potential way to solve this problem is the personalized recommendation system. For example, the well-known e-commerce platform Amazon recommends other products that may be of interest to users by using behavior records such as clicking, browsing, collecting, and adding to shopping carts that can reflect users’ purchasing interests. According to the preferences of each user, intelligent content recommendations based on "thousands of people and thousands of faces" can effectively improve key indicators such as user activity, length of stay, payment rate, retention rate, etc., a...

AI Technical Summary

Problems solved by technology

Recommendations in a single field cannot effectively share Internet resources, resulting in relatively occluded information and easy formation of information islands

Images