A New hybrid Feature selection-Classification model to Improve Cancer Sample Classification Accuracy in Microarray Gene Expression Data

Machine learning techniques are one kind of techniques of Artificial Intelligence that enables systems to learn and improve from data without being explicitly programmed. Machine learning techniques are widely used in medical applications since it has the property to detect inherent patterns from large and complex datasets. Cancer classification based on bio molecular gene expression data is a very crucial topic for medical science as it helps to improve the diagnostic accuracy of cancer samples and is very useful in cancer sample detection and prognosis. But the traditional classifiers performance vitiates due to presence of high feature dimensionality and class imbalance problem present in microarray data. So, in this research work, a new computer aided diagnostic tool is being proposed for cancer sample classification based on bio molecular gene expression data. This tool called MI-TLBO-EB operates in two phases. The first phase selects the best features from the dataset using mutual information and teaching learning based optimization algorithm named MI-TLBO algorithm and the second phase classifies the cancer samples with the help of an extended version of bagging. The proposed model is advantageous in many ways. It helps to curb the curse of higher dimensionality and increases the classification accuracy via handling class imbalance problem with the help of bagging model. The model is applied on different high dimensional microarray gene expression datasets for cancer sample classification and from the experimental results, it has been found that the generalization performance/testing accuracy of the proposed hybrid model is significantly better compared to other well-known existing models.