The application of feature selection methods to analyze the tissue microarray data

In this paper, two feature selection methods, binary genetic algorithm (GA) and sequential floating forward selection (SFFS), were deployed to analyze tissue microarray dataset. The tissue microarray materials in our experiments consisted of 15 tumor-related genes in histological normal tissues adjacent to clinic tumors and different tumors, and the data were arranged in three different datasets and all the collection works were done by the Affiliated Zhongshan Hospital of Xiamen University. For each dataset, we used three distinguished classifiers to obtain the AUC of receive operating characteristic (ROC) curve. The experimental results showed that both feature selection methods could lead to reliable and accuracy results, and be used to discover the connection of genes and cancers.