Recommendation device and method based on user model evolution

Abstract

Disclosed are a recommending device and a recommending method based on user model evolution. The recommending device comprises a user model constructing unit, an evolution mode analyzing unit, and a recommending content generating unit. The user model constructing unit is configured to receive user log information from one or more client devices and construct a user model for each user based on the user log information of the user of each client device. The evolution mode analyzing unit is configured to analyze a time evolution mode of the user model for each user. The recommending content generating unit is configured to predict a user model evolution trend of a target user based on analysis results of the evolution mode analyzing unit, generate a recommending content corresponding to a predicted result, and send the recommending content corresponding to the predicted result to the client device of the target user, wherein the user log information comprises data about user browsing behaviors when the user is using the Internet, and wherein the evolution mode represents the change of interests in different periods of the user.

Description

technical field [0001] The present application relates to a recommendation device and method, more specifically, to a personalized recommendation device and method based on user model evolution. Background technique [0002] With the rapid development of Internet technology, the number of services that can be used on the Internet has increased dramatically, and it has become more and more difficult for people to find interesting information in massive data. Therefore, there is a need for a method that can recommend corresponding content (for example, web pages) to a user when the user searches or browses the web according to different interests and preferences of each user, so as to provide users with targeted personalized recommendation services. [0003] Existing personalized recommendation methods mainly include two types: content recommendation method and collaborative filtering method. The content-based recommendation method is suitable for the situation where the cont...

AI Technical Summary

Problems solved by technology

[0004] However, the existing methods are all based on user modeling based on historical data, and make recommendations based on the user's history and / or existing interest preferences or behaviors, which can only reflect the user's past or existing interest or behavior patterns. without taking into account future trends in user interests or behavior

In addition, the existing personalized recommendation methods consider the time factor for the user modeling process, but do not extend it to the prediction of the user's future interests or behavior trends.

Images