Liver Cancer Detection

Abstract

Machine learning techniques play an important role in building predictive models by learning from Electronic Health Records (EHR). Predictive models building from Electronic Health Records still remains as a challenge as the clinical healthcare data is complex in nature and analysing such data is a difficult task. This paper proposes prediction models built using random forest ensemble by using three different classifiers viz. J48, C4.5 and Naïve Bayes classifiers. The proposed random forest ensemble was used for classifying four stages of liver cancer. Using a feature selection method the reliable features are identified and this subset serves as input for the ensemble of classifiers. Further a majority voting mechanism is used to predict the class labels of the liver cancer data. Experiments were conducted by varying the number of decision trees generated using the J48, C4.5 and Naïve Bayes classifiers and compared with the classification made using decision stump and Adaboost algorithms.