Forgetting characteristic-based user similarity calculation method in collaborative filtering recommendation system

Abstract

The invention relates to a forgetting characteristic-based user similarity calculation method in a collaborative filtering recommendation system. The method comprises the steps of constructing a user-project score matrix; constructing a user-project time matrix; calculating forgetting cycles of users; constructing forgetting index functions; constructing improved forgetting index functions; in combination with the forgetting cycles of the users, enabling values of the improved forgetting index functions in each forgetting cycle window not to be changed, and improving the forgetting index functions to gradient index attenuation functions for the users; calculating project scores of the users and an average score of the users after improvement, wherein the project scores of the users after improvement are products of original scores of a project and corresponding forgetting function values, and the average score of the users after improvement is an average value of all the project scores of the users after improvement; and calculating forgetting characteristic-based user similarity. According to the method, the user interest hobbies can be mastered more accurately and the similarity value reliability is improved.

Description

technical field [0001] The invention relates to a method for calculating user similarity in a collaborative filtering recommendation system. Background technique [0002] In recent years, recommender systems have been widely used in e-commerce (such as Amazon, Taobao, etc.) and some social sites (such as Douban). This further shows that in the rapidly developing Internet environment in the new century, facing massive amounts of data, users need a more intelligent discovery mechanism that can filter useless information and leave only the information they need. [0003] Collaborative filtering recommendation system, as a traditional recommendation system, has gained more favor in applications due to its simple and efficient features, and has also attracted the attention of many researchers. Its core idea is based on the user-item rating data set, screen out users with similar interests to the target user as the nearest neighbor set, and predict the target user's rating of eac...

AI Technical Summary

Problems solved by technology

The present invention provides a user similarity calculation method based on forgetting characteristics. Based on the forgetting characteristic curve of the human brain, the process of information forgetting is regarded as the process of information value attenuation, that is, the attenuation of the reference value of information in similarity calculation. , which also represents the decline of user interests, can more accurately grasp the user's interests and preferences, improve the reliability of the similarity value, and solve the problems caused by data sparseness to a certain extent

Images