Abstract:

The data mining methods used currently are very sensitive to the data being processed, and thus are difficult to be generalized.In order to overcome the shortcomings of the present methods, the de-noise ability of wavelet analysis and the objectivity characteristics of hierarchical clustering analysis (HCA)were integrated by combining the two methods together, and thus a new method—wavelet transform-hierarchical clustering analysis (WT-HCA) was established.The information extracting ability of the new method WT-HCA was investigated.The results showed that WT-HCA could obtain information from metabolomics data effectively.Under the default distance definition, WT-HCA could distinguish the two parental lines totally while HCA could hardly do so.Under another distance definition (distance between samples is the Euclidean distance, distance between groups is the ward distance), the overall percentage of correctly clustered samples after using WT-HCA could reach 93.75%, while HCA could only reach 84.375%.