融合信息瓶颈的模糊三维聚类

摘要 为了有效处理三维列联表数据，采用模糊联合聚类算法的思想，提出一种基于信息瓶颈理论的模糊三维聚类算法（IBFTC）.IBFTC算法为每个维度指定隶属度函数，可实现3个维度上的同时聚类，且在目标函数中引入信息瓶颈理论计算对象与簇之间的距离.采用MovieLens数据集对IBFTC算法进行多方面分析，结果表明，IBFTC算法可获得比现有模糊联合聚类算法更高的聚类准确率.
In order to group three-dimensional data, the thought of fuzzy co-clustering was adopted, and an information bottleneck based fuzzy tri-clustering algorithm, named IBFTC, was presented. The IBFTC specifies membership function for each dimension, simultaneously generates fuzzy clusters on three dimensions and adds information bottleneck theory into objective function for measuring distances between objects and clusters. Experiments on the MovieLens dataset evaluate the performances of IBFTC from several aspects. Experiment shows that IBFTC could achieve higher accuracy than conventional fuzzy co-clustering algorithms.