Microarray Gene Expression Classification for Cancer Detection using Artificial Neural Networks and Genetic Algorithm Hybrid Intelligence

Microarray has been a challenging yet intriguing field in biotechnology. It is also one of the most important approaches that can help diagnose any range of cancers. Microarray gene expressions itself might consist of tremendous amount of genetic codes which can be hard to analyze, but then again only small quantity of the gene expression itself that are worth and effective enough to be conducted in particular cancer diagnostic test. In response to the high dimensionality of the microarray dataset, this research implements principal component analysis (PCA) technique to select the most relevant features. Furthermore, this paper also proposes using artificial neural networks (ANN)-genetic algorithm (GA) Hybrid Intelligence for cancer identification. While ANN is recognized as one of the methods to classify the microarray data, GA in this case is used to optimize the ANN architecture. In order to see the optimization result, ANN-GA model is compared with the ANN model without GA. However, the results of the classification using ANN-GA is able to lift up the average accuracy up to 79.25% and average area under the curve (AUC) value up to 0.7285 than ANN alone. Furthermore, ANN-GA model is able to adjust the predictive accuracy up to 83.33%, 76.47% and 89.93% for Colon Tumor, Prostate Tumor and Lung Cancer respectively.