Prediction of Cancer using Microarrays Analysis by Machine Learning Algorithms

Abstract

The analysis of microarrays that contain information on biomolecules related to different types of cancer is one of the current issues in international scientific research due to the impact it has on public health worldwide. The advances in this scientific research route have been impressive; the different international research groups have applied sophisticated algorithms for machine learning, data mining and related branches with the aim of finding solutions to this problem. The present article contains a study of several the classification algorithms used in the literature, and their application for the prediction of cancer using microarrays analysis. More in detail, we tested six classification models, over microarrays data. The application of the supervised classification algorithms was done over the Weka 3 Software environment, using the Leave One Out validation scheme. In addition, a nonparametric statistical test (the Friedman test) identified the significant differences in the performance of the algorithms, according to the experimental results obtained. The analysis of the hypothesis tests of the experimental results indicates that the Support Vector Machine models outperform others for the prediction of cancer.