基于模糊数学的高维稀疏数据聚类统计方法设计

Design of Clustering Statistics Method for High-dimensional Sparse Data based on Fuzzy Mathematics

传统的数据聚类统计方法仅适用于低维数据聚类问题,为此,设计了基于模糊数据的高维稀疏数据聚类统计方法,以期提升高维稀疏数据的聚类统计效果.以模糊C均值聚类算法为基础,通过优化初始聚类中心解决局部最优问题,缩短聚类统计时间;然后引入权重机制,令该方法适用于高维稀疏数据聚类统计.基于此,以余弦距离替换原有的欧几里德距离,提高高维稀疏数据聚类统计效果.实验证明:在数据维度不同时,该方法均有较优的聚类统计效果.当数据维度较低时,分块比例为10%时聚类统计效果最优;当数据维度较高时,分块比例为40%时聚类统计效果最优.在不同稀疏度等级时,该方法的命中率和聚类统计效率均较高.

Traditional data clustering statistics method is only applicable to low dimensional data clustering problem,therefore,this study designed a high-dimensional sparse data clustering based on fuzzy data statistical method,the clustering of high-dimensional sparse data statistics results.Based on the fuzzy C-means clustering algorithm,by optimizing the initial clustering center,solve the problem of local optimum,shorten the clustering statistics time;Then weighting mechanism are introduced,the method is suitable for high-dimensional sparse data clustering statistics.Based on this,in order to replace the original Euclidean distance,cosine distance to improve the effect of high-dimensional sparse data clustering statistics.Experiments show:the data dimension is not at the same time,this method has a better clustering effect of statistics.When data dimension is low,partitioned clustering statistics result when compared with 10%of the optimal;When high dimension data,block ratio is 40%when the optimal clustering statistics effect.In the sparse degree of different grade,the shooting and cluster statistical efficiency of the method are high.