User browsing behavior based personalized recommendation method and apparatus

Abstract

The invention discloses a user browsing behavior based personalized recommendation method and apparatus. The recommendation method comprises: obtaining browsing history of a user; analyzing the interests and demands of the user according to each keyword in the browsing behavior history of the user, and realizing dynamic update of user interest and demand through a self-detection function of a keyword counter; and creating a list of recommendation content based on user browsing behaviors, and pushing the recommendation content to the user in a non-disturbing manner through an email. Therefore, the personalized content can be recommended according to the user browsing behaviors.

Description

technical field [0001] The invention relates to a personalized recommendation method and device based on user browsing behavior, which realizes information recommendation and belongs to the field of electronic information. Background technique [0002] With the explosive growth of information resources on the Internet, the difficulty for users to obtain the resources they are interested in from the Internet has also increased accordingly. How to enable users to efficiently find the information they are interested in has become a hot spot at present, so personalized recommendation systems came into being. [0003] The current mainstream recommendation algorithms are mainly divided into two categories, one is recommendation based on external correlation algorithms, and the other is recommendation based on content. The former is realized through modeling based on the similarity of interests between users, such as collaborative recommendation. The principle is Use an algorithm ...

AI Technical Summary

Problems solved by technology

[0003] The current mainstream recommendation algorithms are mainly divided into two categories, one is recommendation based on external correlation algorithms, and the other is recommendation based on content. The former is realized through modeling based on the similarity of interests between users, such as collaborative recommendation. The principle is Use an algorithm to find multiple neighbors that are most similar to the target user, and use these multiple neighbors to make recommendations to the target user, but this recommendation method can only be realized based on user evaluations. If a certain content has not been evaluated by any user Or there is no evaluation system in the system at all, then this item cannot be recommended, so this kind of method has great limitations and cannot be applied to information systems without evaluation functions

Other such algorithms also have the problems of low efficiency, large system resource occupation, large amount of recommended information, and low accuracy. The latter analyzes the internal structure and semantic information of the items to find out certain items that the target users may be interested in. In order to make recommendations, however, most of the known content recommendation methods currently adopt the method of user-defined keywords, without too complicated algorithms, occupying relatively less system resources, and improving the accuracy rate. However, in reality, user needs will change according to space and time. Under this premise, if you want to accurately recommend information for users, you can only manually change the predetermined keywords, which will lead to cumbersome settings. Moreover, if the description of the predefined keywords is inaccurate, the recommendation information will be incorrect and the meaning of the recommendation will be lost. In addition, the above two current methods all use the instant recommendation method to directly display the results of the recommendation information on the user's browsing page. On the interface, this method does not respect the user's behavior, which not only affects the user experience, but also affects the user's privacy

Images