Project recommendation method, apparatus and device based on recommendation system

Abstract

The invention discloses a project recommendation method based on a recommendation system. The method includes: calculating the preference of each user for each project according to the scoring data ofeach user for each project and the average scoring data of each project; integrating each preference value with the corresponding scoring data to obtain the association weights of each user for eachproject; building user-project bigraph between users and projects based on association weights; based on the random walk algorithm for the user-project bigraph, obtaining the first recommended projectlist by random walks in the second part of the project diagram. The invention comprehensively considers the overall score of the project and the preference degree of the user to make recommendation,so that the finally obtained first recommendation project list not only includes hot projects, but also includes projects which are partially preferred by the user but have low overall score. Therefore, the diversity of project recommendation is improved.

Description

Technical field [0001] The present invention relates to the field of recommendation systems, in particular to an item recommendation method, device and equipment based on the recommendation system. Background technique [0002] With the development of the Internet, more and more data are generated on the Internet, and the problem of "information overload" has appeared. For users, many users need an effective information filtering method to filter out the content that meets their interests from the massive network data. As an important means, recommendation systems have been widely used in various commercial websites and Internet-related fields. It helps users find items or services they are interested in in a large amount of network data. [0003] The current recommendation system is basically aimed at improving the accuracy of recommendation, and the recommendation results basically tend to recommend popular items. The popular items here refer to the items with high overall user...

AI Technical Summary

Problems solved by technology

[0003] The current recommendation system basically aims to improve the accuracy of recommendation, and the recommendation results basically tend to recommend popular items. The popular items here refer to the items with high overall ratings of users, such as the top items with high average ratings of all users. N items, etc., but such a recommendation result does not mean that the user satisfaction is high, because if the recommendation results are concentrated on popular items, the number of popular items is small, and there is a large similarity between some popular items ( This is because related items of certain popular types will all belong to popular items), which will lead to insufficient diversity of recommendation results

That is to say, there is a long-tail effect in the current recommendation results: most of the high scores are concentrated on a small number of popular items, while most of the remaining unpopular items are rarely recommended, which makes it difficult for users to know about these unpopular items, thus As a result, its ratings are getting less and less, making it harder to be recommended

Images