Cancer Detection Based on Microarray Data Classification with Ant Colony Optimization and Modified Backpropagation Conjugate Gradient Polak-Ribiére

Abstract

Based on IARC, cancer is the deadliest disease in the world. Microarray data technology is created to make it easier for doctors to diagnose cancer faster. This technology brings a glimmer of hope for researchers to prevent cancer from an early age. Microarray data has huge data dimension, with hundreds of sample and thousands of features. This paper presents a classification system using Modified Backpropagation with Conjugate Gradient Polak-Ribiere and Ant Colony Optimization as the gene selection. By using the fundamental function of human body’s neural network, MBP Conjugate Gradient Polak-Ribiere can classify the microarray data, whereas, with the application of ACO as gene selector, important genes will be selected so that MBP optimization is achieved. MBP has been known for its ability to process microarray data with huge dimension. MBP is perfect for microarray data processing. While ACO is a new method developed by previous researchers to perform feature selection. In this study, it is found that the classification of MBP can reach the F-Measure score of 0.7297. When combined with ACO as feature selection, the score increases by 0.8635. ACO is proven to optimize the classification method of microarray cancer data very well.