Cancer Detection Based on Microarray Data Classification using Genetic Bee Colony (GBC) and Conjugate Gradient Backpropagation with Modified Polak Ribiere (MBP-CGP)

Abstract

Cancer is one of the major health problems in the world, and should therefore be detected as early as possible. The development of technology has given rise to microarray technology, which can help researchers gather information from thousands of genes in a human being simultaneously, which is useful for the detection of cancer. Each feature of microarray data has a high dimension, so dimensional selection is done to improve the accuracy of microarray data classification; the Genetic Bee Colony (GBC) algorithm and Conjugate Gradient Backpropagation with Modified Polak Ribiere (MBP-CGP) can be used to detect whether or not an individual has cancer. GBC is a metaheuristic hybrid algorithm based on the Artificial Bee Colony (ABC) algorithm and Genetic Algorithm. MBP-CGP is a modification of the Artificial Neural Network (ANN), designed to accelerate backpropagation training. By implementing GBC and MBP-CGP as the feature selection method and classifier, respectively, the system is able to select features of up to 47-51% for all datasets with the performance generated for all datasets (without GBC) ranging between 63.75-84.44% for the MBP-CGP architecture with two hidden layers and 63.75-82.77% for the MBP-CGP with one hidden layer. Meanwhile, the accuracy of results using MBP-CGP and GBC classifications ranged between 88.75-100% for all datasets with one hidden layer.