基于局部敏感哈希的隐私保护实时服务推荐

Real-time Service Recommendation with Privacy-preservation Based on Locality Sensitive Hashing

协同过滤算法是服务推荐系统中最有效和应用最广泛的推荐方法,其侧重于提高推荐结果的准确性。然而,在大数据背景下,用户行为数据不仅经常频繁更新而且数据规模增长迅速,传统的协同过滤算法需要穷举搜索所有数据,相似度计算耗时较高,推荐效率低,无法满足用户实时体验的需求服务。快速从大数据中获得高质量的推荐服务成为一种新的需求,为此,提出基于局部敏感哈希技术的协同过滤算法,算法过滤了绝大多数不相似的项目,避免了冗余的相似度计算,另一方面算法将用户行为数据哈希为二进制哈希编码,进而保护用户隐私。最后,在不同规模尺寸的数据集上与主流算法对比,实验表明提出的算法在效率和准确度间能够取得较好的折衷。

Collaborative filtering algorithm is the most effective and widely used recommendation method in service recommen⁃dation system,which focuses on improving the accuracy of recommendation results.However,in the context of big data,user behav⁃ior data is not only frequently updated,but also the data scale is growing rapidly.The traditional collaborative filtering algorithm ex⁃haustive search over all of the data,which can not meet the needs of users'real-time experience due to high time cost of similarity calculation and low efficiency.It has become an emerging demand to quickly obtain high-quality recommendation services from big data.Therefore,a collaborative filtering algorithm based on local sensitive hashing technology is proposed.On the one hand,the al⁃gorithm filters most dissimilar items and avoids redundant similarity calculation.On the other hand,the algorithm hashes user be⁃havior data into binary hash codes,so as to protect user privacy.Finally,compared with the mainstream algorithms on datasets of different scale,experiments show that the proposed algorithm can achieve a good compromise between efficiency and accuracy.