基于改进局部敏感哈希的协同过滤推荐算法

Collaborative Filtering Recommendation Algorithm Based on Improved Local Sensitive Hashing

传统推荐系统中存在用户评分数据高维稀疏、分布不均匀和传统用户相似度计算准确性低等问题,本文提出一种基于改进局部敏感哈希的协同过滤算法。首先利用改进局部敏感哈希算法对用户评分数据进行降维处理并构建索引,并使用相似度修正系数对用户相似度计算做出改进;然后利用索引敏捷切确地计算目标对象的近邻用户集合;之后选择近邻用户聚集的高相似度用户,使用加权算法对目标对象未评分项目进行评定预估。实验结果表明,对于非均匀用户评分数据的高维稀疏问题,该算法不仅能明显缩短近邻用户检索时间,且能有效提高推荐精度。

A collaborative filtering algorithm based on improved locality sensitive hashing made to solve the problems of large scale,uneven distribution of high-dimensional sparse user scoring data and the calculation of traditional user similarity with low accuracy in recommendation system.Firstly,the algorithm reduced the dimension of user rating data and build index for it.The calculation of user similarity is improved by similarity correction coefficient.Then the index is used to calculate the nearest neighbor user set of the target user quickly and accurately.Finally,the algorithm selected the highly similar users in the nearest neighborhood,and use the weighted strategy to predict score for the target users,so as to achieve collaborative filtering recommendation.The results of the trial show that the algorithm can effectively solve the problem of high-dimensional sparsity of uneven user rating data,and it not only shorten the retrieval time of nearest neighbor users,but also effectively improve the recommendation accuracy.