基于改进预测强度的大数据K-均值聚类方法

A Large Data Clustering Method Based on Improved Prediction Strength

为了降低偶然因素的影响,提出了一种基于改进预测强度的大数据K-均值聚类方法,其基本思想是:首先将数据集若干等分,每一等分轮流作为测试集,取其平均预测强度,然后根据预测强度确定聚类数和聚类变量,再用K-均值聚类方法对数据集进行聚类。用上述方法研究了访客在某网站各栏目的平均停留时间,结果表明,基于预测强度的聚类方法较常规聚类方法更适宜于大数据的聚类分析。

In order to reduce the influence of accidental factor,a large data K-means clustering method based on improved prediction strength is put forward.The basic idea of method is that first data set is divided into equal parts,and each part is set up test set in turn.The average strength prediction is computed,and clustering number is determined according to the strength prediction,then K-means clustering method is applied for data set.By means of the above method,the average residence time of the visitors in a website is studied.The results show that the clustering method based on the prediction strength is more suitable for the cluster analysis of large data.