Classification of genetic expression in prostate cancer using support vector machine method

Abstract

Prostate cancer has long been a concern of expert’s human genetics in health research. However, an explanation of the main causes of prostate cancer cannot be obtained metabolically-biologic, except the most common one of which is heredity. Explanation of the risk of contracting prostate cancer is sought through genetic explanation of prostate cancer cells and healthy prostate cells from DNA sequencing in the form of micro arrays data or in the form of Gleason values. Cancer cell genetic data is high dimensional where the number of variables observed were far more than the individual observed. It’s make ordinary multivariate classification techniques fail to handle this data because of the singularity matrix. In addition, the observations number of cancer patients are small since they are rarely found. With these two facts, then in this paper we will use a machine learning approach to study the classification, namely SVM. SVM will be compared with the Naive Bayes Classifier and Discriminant Analysis method to determine the accurate division in distinguishing prostate cancer cells from healthy prostate cells. The sample data used consisted of 102 people with 2135 genetic variables which were then divided into training data and testing data. Based on the results of the study, the classification by the SVM method has an accuracy value of 96% with a precision error in the tumor class of 7%. The Naive Bayes classification has a precision error of 23.5% with a classification accuracy of 84%. While the Discriminant Analysis method produces an accuracy of 92% with a precision error of 13.33%.