动态自适应的混合智能协同推荐算法

针对当前协同过滤推荐算法存在数据稀疏、用户兴趣变化和时效性不明显、推荐质量差等问题,提出了一种动态自适应的混合智能协同过滤推荐算法。首先利用修正核模糊聚类算法进行聚类分析,得到目标用户初始邻居集,缩小计算范围；重新定义了初始等价关系和等价关系相似性,提出了动态x近邻算法,得到准确邻居集并用预测评分填充矩阵,优化数据质量；最后引入用户兴趣变化因子和评价时效,挖掘用户潜在的兴趣变化,得到较好的推荐结果。实验结果表明,该算法能够得到更准确的最近邻居集,提高预测准确率和推荐质量,为用户提供更好的个性化推荐。

Dynamically adaptive hybrid intelligent collaborative filtering recommendation algorithm

In order to solve the problems of current collaborative filtering algorithm,such as sparse data, inconspicuous user interest changes, timeliness and poor recommendation quality, an adaptive hybrid intelligent algorithm was proposed. The initial neighbor set of the target user was got by modified kernel fuzzy clustering analysis firstly, which reduced the calculation range; furthermore, the initial equivalence relation and equivalence relation similarity were redefined, and a dynamic x nearest neighbor algorithm was proposed to get the accurate neighbor set, and then to fill the matrix using the prediction score, which optimized score data quality. At last, the interest change factor and rating time weight of the users was introduced, and mined potential interest changes to obtain better recommendation. The experimental results show that the algorithm can get more accurate nearest neighbor set, which can improve the prediction accuracy and the quality of recommendation, and provide better personalized recommendation for users.