Microarray Gene Expression for Cancer Classification using Fast Extreme Learning Machine with ANP

229 | P a g e Abstract—Cancer is a class of diseases in which a set of cells display uncontrolled growth, invasion that interrupts upon and demolishes nearby tissues, and occasionally metastasis, or spreading to other locations in the body via lymph or blood. Cancer has become one of the dangerous diseases in the present scenario. DNA microarrays turn out to be an effective tool utilized in molecular biology and cancer diagnosis. Microarrays can be utilized to determine the comparative amount of particular mRNAs in two or more tissue samples for thousands of genes concurrently. As the supremacy of this technique has been identified, various open queries arise about suitable examination of microarray data. The multicategory cancer classification is playing a vital role in the field of medical sciences. As the numbers of cancer victims are increasing steadily, the necessity of the cancer classification techniques has become indispensible. For the above impenetrability and to obtain better consequences of the system with accuracy a new learning algorithm called Extreme Learning Machine (ELM) is used. ELM overcomes difficulties such as local minima, inappropriate learning rate and over fitting usually occurred by iterative learning techniques and performs the training rapidly. In this approach, the performance of the ELM is improved through the use of Analytic Network Process (ANP) and Levenberg Marquardt training algorithm. This approach also utilizes the error free ANOVA techniques in the preprocessing stage. This paper represents that ANOVA technique can be utilized to normalize microarray data and afford determination of alterations in gene expression that are corrected for potential perplexing effects. The proposed technique is evaluated with the help of Lymphoma data set. The experimental result represents that proposed technique results in better classification accuracies with lesser training time and implementation complexity compared to conventional techniques. Keyword--ELM, ANOVA, Cancer Classification and Gene Expression, Fast ELM, ANP.