According to the data from the World Health Organization (WHO) in 2012, cancer is considered as the leading cause of death in the world. About 8.2 million people died because of cancer and the number is estimated to increase each year due to an unhealthy lifestyle [15]. Deaths due to cancer could be prevented if it is detected early. In recent decades, microarray has taken an important role in cancer research. Microarray is a technology that is capable of storing thousands of gene expressions taken from several specific tissues of human at once. By analyzing microarray data, it can be known whether the tissues are affected by cancer or not. This study provides a fast and accurate framework for cancer detection based on microarray data classification using principal component analysis (PCA) and modified back propagation (MBP). MBP is a modification of standard back propagation (BP) that implements conjugate gradient algorithm on search direction in BP training. The experiment results show that the proposed system (MBPorPCA+MBP) is able to outperform BP-based system (BP or PCA+BP) in accuracy and especially in training time.

cancer detection, microarray data, modified back propagation, principal component analysis.