A Cross-Correlation Collaborative Filtering Method Integrating Users' Weak Trace Behavior Preferences

Abstract

The invention discloses a cross-correlation collaborative filtering method fusing weak trace behavior preferences of a user. According to the method, weak trace behaviors such as clicking and articlebrowsing time are collected, a scoring matrix and a weak trace behavior matrix of the user are constructed, an arithmetic scoring mean value of the user is assigned to represent weak trace behavior preferences of the user, and a scoring enhancement matrix fused with the weak trace behavior preferences of the user is obtained; the correlation between the user behavior preferences and articles can be described more accurately; furthermore, an improved cross correlation coefficient calculation method is provided, a hot penalty coefficient is fused to construct a user-user preference similarity matrix, the preference similarity of the user to the articles is predicted, and recommendation sorting of the articles with similar preferences is optimized. According to the method, the problem of datasparseness in a traditional collaborative filtering method is solved, and the influence of factors such as different user scoring scales and non-preference behaviors caused by popular factors on userpreference similarity calculation is reduced, so that the accuracy of similar preference recommendation results and sorting is effectively improved.

Description

technical field [0001] The invention belongs to the field of intelligent information recommendation, and in particular relates to a cross-correlation collaborative filtering method that integrates users' weak trace behavior preferences. Background technique [0002] The rapid development of the Internet has brought a huge amount of information, and the time users spend on information retrieval is increasing day by day. Information intelligent recommendation can analyze user preferences and potential interests based on the user's historical behavior, and perform data mining and intelligent recommendation on information, so as to meet the individual needs of users and save users time for information screening. The traditional collaborative filtering recommendation method is a common method in the field of information intelligent recommendation, which can be divided into two categories: user-based collaborative filtering and item-based collaborative filtering: user-based collab...

AI Technical Summary

Problems solved by technology

The sparsity of the "user-item" scoring matrix in the traditional collaborative filtering method has always been the main problem affecting the recommendation accuracy, especially with the exponential growth of the number of Internet users and the number of items, the sparsity problem has become more prominent; at the same time, the user rating scale Non-preference behaviors caused by different and popular factors also cause deviations in the calculation of user preference similarity, increase the difficulty of sorting similar user preferences, and bring new challenges to traditional collaborative filtering methods

Images