基于变异系数的SOM算法在多维分析中的应用研究

Research on SVM Algorithm Based on Variation Coefficient in Multidimensional Analysis

在多维分析领域,自组织映射SOM算法是一种无导师学习方法,具有降维,自组织的,可视化等特性,由于SOM算法计算获胜神经元采用欧式距离的原因,忽略了不同维度对于相似度的贡献,针对该不足,该文采用变异系数对维度权重进行研究和改进SOM算法。实践证明:相比没有维度权重的SOM算法,采用带有指标权重的SOM算法具有更好的准确率和凝聚度。

In the field of multidimensional analysis, the self-organizing mapping SOM algorithm is a kind of no-tutor learning method, which has the characteristics of dimension reduction, self-organization and visualization. Because the SOM algorithm calculates the distance of the winning neurons using Euclidean distance, This paper uses the coefficient of variation to study the dimension weight and improve the SOM algorithm. It is proved that compared with the SOM algorithm without dimension weight, the SOM algorithm with the index weight has better accuracy and cohesion.