Feature Selection For High-dimensional Cancer Protein Mass Spectrometry Data

In this paper, the main research the high-dimensional feature extraction of protein spectrum data problem. Since the birth of spectrum technology, its application is more and more widely, also is extremely important in the field of protein research. Protein spectrum in the field of cancer identification, wide application. However, protein spectrum data of research has been beset with high dimension problems. In the process of protein spectrum data research, as the growth of the amount of data and data dimension, how to extract the data dimension reduction with the effective feature is more and more important. This paper in view of the higher dimensional protein spectrum cancer data in the process of dimension reduction problem, put forward based on low frequency coefficients of wavelet analysis and principal component analysis(and another based on high frequency coefficients of wavelet analysis and principal component analysis)of high-dimensional protein spectrum cancer data feature extraction method, and after feature extraction, using support vector machine (SVM) and linear discriminant to classify the data.We use wavelet decomposition on 8-7-02 data set at second level,use different wavelet basis(db1,db3,db4,db6,db8,dbl0,haar) and classify them with support vector machine,then we get different recognition rates:98.18%,98.35%,98.04%,98.36%,97.89%,97.96%,98.20%.The method remain or improve the classification accuracy and reduce the work time compared with the methods of previous researchers.