Commodity information recommending method and commodity information recommending system based on user historical behaviors

Abstract

The invention discloses a commodity information recommending method based on user historical behaviors. The method comprises the following steps of S11, acquiring historical behavior data of a user in an E-commerce website; S12, establishing a user commodity probability prediction characteristic vector according to the historical behavior data; S13, training a model according to the user commodity probability prediction characteristic vector, and obtaining a user recommendation commodity prediction model; S14, inputting to-be-predicated user data into the user recommendation commodity prediction model, and calculating a predicated purchasing probability of a behavior commodity; and S15, according to the predicated purchasing probability, calculating the predicated purchasing probabilities of correlated commodities, and combining the behavior commodities and the correlated commodities for obtaining a commodity recommending list. According to the commodity information recommending method and the commodity information recommending system, behavior data of the user are accurately analyzed; an individualized commodity recommending list is supplied to a user; and furthermore higher accuracy in commodity recommending is realized.

Description

technical field [0001] The present invention relates to a commodity information recommendation method and system in the field of electronic commerce, in particular, to a commodity information recommendation method and system based on user history behavior. Background technique [0002] The methods of using e-commerce user behavior to recommend products include: rankings based on user behavior (such as sales rankings), recommendations based on the same behavior (for example, when everyone saw product A at the same time, what other products did they see), and rankings based on the same theme Recommendations (such as children's mobile phone recommendations), recommendations based on user feedback (such as products that everyone likes), etc. By asking questions, customers answer questions, directly understand customer preferences, and recommend suitable products. In addition, there is a product recommendation method based on association rule equations. [0003] The above metho...

AI Technical Summary

Problems solved by technology

These methods are all aimed at predicting the purchase probability of users or users for categories, and cannot predict the purchase probability of unpurchased products or single items.

Images