改进Parzen窗解决高维数据聚类的方法研究

由于高维数据聚类的现实意义日益增强，而Parzen窗估计法仅对低维数据集聚类能获得良好的结果，随着维数增加，效率降低，因此对Parzen窗进行加权改进，通过多次仿真实验确定加权函数，将高维数据投射至低维空间，对其聚类，逐步投向高维空间，对结果矩阵进行优化处理，得到更为优良的聚类效果。

Study of clustering using improved Parzen window on high-dimension dataset

Due to the realistic meaning of clustering on high-dimension dataset,excellent result can be acquired when clustering dataset using traditional Parzen window on low-dimension.With the higher dimension,the efficiency decreases quickly.Through many simulate experiments,the weighting function is gained.The high-dimension dataset is shadowed on lower-dimensional space,and clustered,then shadowed back to higher-dimensional space.The result matrix is optimized,and the better clustering effect is gained.