A Study on Feature Selection Method based on Random Forest for Cancer Diagnosis System
Gyoo Seok Choi, Jong jin Park, Ha-Nam Nguyen · The Journal of Korean Institute of Information Technology · 2008
One of the most important problems in bioinformatics is how to extract the useful information from a huge amount of data, and make a decision in diagnosis, prognosis, and medical treatment applications. Machine learning approaches such as Neural network, Decision tree, Support Vector Machines are well suited for domains characterized by the presence of large amount of data, noisy patterns, and absence of general theory. The main goal of our research is to propose an efficient feature selection method to achieve a cancer diagnosis system with high accuracies, and good adaptability to clinical dataset. We propose a new feature selection method based on Random Forest(RF) to be applicable to the learning algorithm for Cancer Diagnosis System. The experiments on clinical dataset such as colon cancer indicate that our proposed method obtain higher and more stable classification performance than the baseline methods. Our method also results in comparable and better classification performance than other classification methods.