Application of improved SOM network in gene data cluster analysis

Abstract

At present, cluster analysis has become a very good channel for analyzing gene expression data to obtain biological information. In recent years, many experts have used traditional clustering algorithms and new clustering algorithms to mine gene expression data. This article first introduces the preprocessing of gene expression data. Then, by using principal component analysis (PCA) to process the gene data, a small number of characteristic variables are extracted as new indicators, and the indicators are evaluated to achieve the purpose of dimensionality reduction. The dimension reduction index is applied to the dynamic self-organizing neural network (DSOM) neural network, and the victory neurons are selected by the minimum Euclidean distance. The characteristics of the genetic data are clustered through the DSOM network, and the gene types with similar characteristics are divided into one region. The results show that PCA and DSOM networks have a high accuracy rate for clustering of genetic data, and a clear division of boundaries.