基于中心平面的聚类模型及在电商中的应用

因为k-平面聚类算法(kPC)和k-中心平面聚类算法(kPPC)构建的聚类中心平面是无限延伸的,这会影响聚类的性能,所以提出了局部的k-中心平面聚类(L-kPPC)算法.此算法在kPPC中引入了k-均值聚类算法(k-mean),这样使得样本点都聚集在类中心周围.L-kPPC利用了各聚类中心平面的局部特征构建类中心平面,使同一类的数据点到此类的聚类中心或平面尽可能的近,离其他类中心或平面尽量远,这导致求解特征值问题.在此,利用拉普拉斯图建立初始化的数据点,而不是随机选择的初始数据点.最后从电商平台ebay提供的Web Service接口提得数据进行实验,实验结果分析表明,L-KPPC算法有较好的表现.

Research on Clustering Algorithm Based on Central Plane

Because the center plane of the k-plane clustering algorithm(kPC)and the k-center plane clustering algorithm(kPPC)is infinitely extended,this will affect the performance of the cluster,so a local k-center plane clustering(L-kPPC)algorithm is proposed.This algorithm introduces the k-mean clustering algorithm(K-mean)into kPPC,so that all the sample points are gathered around the class center.L-kPPC uses the local features of each cluster center plane to construct the center plane of the class,so that the same class of data points are as close as possible to such cluster centers or planes,as far as other classes center or plane are as far as possible,which leads to the solution of the eigenvalue problem.Here,we use Laplasse diagram to establish initialized data points instead of randomly selected initial data points.Finally,experiments are carried out from the Web Service interface provided by eBay,and experimental results show that L-KPPC algorithm performs well.