An Efficient Machine Leaning based Gene Expression Cancer Diagnosis

Abstract

Diagnosis of cancer became an active area by using gene expression profiling based classification. Many researches contributed their approach towards it but direct multiclass classification is much more difficult than binary classification and the classification accuracy may drop dramatically when the number of classes increases. Due to that they use combination of binary classifier on One-Versus-All (OVA) or a One-Versus-One (OVO). In this paper Extreme Learning Machine (ELM) is used for directing multi category classification problems in the cancer diagnosis area. ELM avoids problems like local minima, improper learning rate and over fitting commonly faced by iterative learning methods and completes the training very fast. To evaluate the result this proposed method is compared with SVM-OVO method using performance metrics such as Averaged Number of Hidden Nodes (Support Vectors), Accuracy and Evaluation time conforms that this method suits well for Microarray Gene Expression Cancer Diagnosis classification.