基于兴趣相似度传递的增强LSH统计预测算法

随着在线用户和物品数量的不断增长,有必要通过追踪和筛选历史数据,为用户提供机制可参考的决策建议。构建统计预测算法是实现启发式预测用户兴趣的有效机制。因此,在充分利用用户自身历史偏好和潜在偏好的前提下,提出兴趣相似度传递思想,分析用户的社交关联强度,计算用户的邻近社交兴趣和选择趋向特征,设计并实现了可扩展的局部敏感哈希(Improved Local Sensitivity Hashing,ILSH)统计预测算法。实验表明,该算法在有利于相似度计算量剧增的背景下,在提高兴趣预测的准确性和可靠性方面优于其他近似算法。

AN IMPROVED LSH STATISTICAL PREDICTION ALGORITHM BASED ON INTERESTED SIMILARITY PROPAGATION

As the number of online users and items continues to grow,it is necessary to track and sift through historical data to provide mechanisms for users to make informed decisions.Constructing statistical prediction algorithm is an effective mechanism for heuristic prediction of user interest.Therefore,on the premise of making full use of the user s own historical and potential preferences,this paper proposes the idea of interest similarity transfer,analyzes the intensity of user s social association,calculats the user s neighboring social interest and the choice of trend features,and designs and implements ILSH(improved local sensitivity hashing)statistical prediction algorithm.Experiments show that the proposed algorithm is better than other approximation algorithms in improving the accuracy and reliability of interest prediction in the context of increasing similarity computation.