Improved hybrid collaborative filter recommendation method

Abstract

The invention discloses an improved hybrid collaborative filter recommendation method, and aims at remitting influences of data sparsity, improving the recommendation accuracy and supplying the gap that two single collaborative filter algorithms cannot give consideration to correctness and personality. The method comprises the following steps of: weighting a traditional similarity calculation manner by combining quantity difference of common grading items, optimizing a similarity result and enabling a construct interest model of neighbor sets to be closer to a target user and a target project;importing a concept of neighbor set similarity quality to measure similarity levels of the neighbor sets, deciding proportions in a hybrid model on the basis of user collaborative filter and projectcollaborative filter, and importing a control factor to improve the influences of the data sparsity so as to optimize a grading prediction result.

Description

technical field [0001] The invention belongs to the field of recommendation system and data mining, in particular to a hybrid collaborative filtering technology based on the similarity quality of users and items. Background technique [0002] With the advent of the progress of the information society, the amount of information data that users can obtain is increasing, and the problem of information overload is becoming more and more serious. How to quickly obtain the information they need from these massive information data has become a big data era. hot topics. The emergence of recommendation algorithms has changed the way users interact with information data: from actively obtaining information to actively recommending information to users. The focus and difficulty of the recommendation algorithm lies in how to effectively improve the accuracy of information recommendation. [0003] Collaborative filtering algorithm is the most widely used recommendation algorithm. It do...

AI Technical Summary

Problems solved by technology

When the scoring matrix is ​​sparse, it is difficult for the algorithm to accurately find the neighbors of the target user or the target item. When building the neighbor set, some information will be lost, which will cause deviations in subsequent score predictions and lead to a decline in recommendation quality.

[0006] (2) Cold start problem

When a new item appears, there is no user to evaluate it, and the traditional collaborative filtering algorithm cannot predict and recommend it; Unable to recommend item to this user

[0007] (3) Algorithm defects

The traditional similarity calculation method does not take into account the impact of the difference in the number of common scoring items on the construction of the neighbor set; at the same time, a single collaborative filtering algorithm can only consider the interaction of user information or item information, ignoring the interaction between the two The impact on score prediction cannot take into account both accuracy and personalization

Images