Multiclass classification of leukemia cancer data using Fuzzy Support Vector Machine (FSVM) with feature selection using Principal Component Analysis (PCA)

Abstract

Cancer is the second leading cause of death globally. According to WHO prediction (2015) cases of cancer deaths will increase to 21.6 million cases by 2030. Therefore, early detection of cancer is necessary to avoid the spread of cancer and machine learning is required to increase performance in the detection of cancer. In general, microarray cancer data consist of many features. However, there are several features in cancer data that did not have important information in classification cancer. Therefore, these features will be summarized from several features under some common underlying factors into fewer components using the Principal Component Analysis (PCA) method. Then, we select the most features who have important information for classification cancer. This paper focuses on the comparison of using and without the PCA method on cancer data coupled with the Fuzzy Support Vectors Machines (FSVM) method for cancer classification. The experimental results, without the PCA method on cancer data coupled with the FSVM method for cancer classification the accuracy is 87.69 % and by using the PCA method on cancer data coupled with the FSVM method for cancer classification the accuracy is 96.92 % (obtained by using 60 features).