Machine Learning Techniques for Cancer Risk Prediction

Abstract

Microarray data analysis plays a vital role in cancer classification and diagnosis. But it’s a big challenge to achieve high level of accuracy in cancer classification with large set of genes. The totalcounts of features areregularlygreater than that of the number of instances. For this reason, it needs to achieve feature selections for organisation of genes. Feature collection lowers theproblematicthrough selecting informative features from datasets. In this study, four feature selection methods with ranker search techniques tool of Weka are used to select top 100 informative genes, the classification technique Support Vector Machine, Random Forest, Random Tree are applied to these selected genes to conduct experimental work on the presented data-sets. The experimentally result that the projected feature assortment and lengthfall in data volume stretchesimproved result of accuracy.

Keywords:Cancer Classification, Feature Selection, Microarray data, SVM, Random Forest, Random Tree