Liver Disease Prediction Model Based on Oversampling Dataset with RFE Feature Selection using ANN and AdaBoost algorithms

Abstract

Liver disease counts are one of the most prevalent diseases all over the world and they are becoming very common these days and can be dangerous. Liver diseases are increasing all over the world due to different factors such as excess alcohol consumption, drinking contaminated water, eating contaminated food, and exposure to polluted air. The liver is involved in many functions related to the human body and if not functioned properly can affect the other parts too. Predication of the disease at an earlier stage can help reduce the risk of severity. This paper implemented oversampling dataset, feature selecting attributes, and performance analysis for the improvement of the accuracy of classification of liver patients in 3 phases. In the first phase, the z-score normalization algorithm has been implemented to the original liver patient data-sets that has been collected from the UCI repository and then works on oversampling the balanced dataset. In the second phase, feature selection of attributes is more important by using RFE feature selection. In the third phase, classification algorithms are applied to the data-set. Finally, evaluation has been performed based upon the values of accuracy. Thus, outputs shown from proposed classification implementations indicate that ANN algorithm performs better than AdaBoost algorithm with the help of feature selection with a 92.77% accuracy

Keywords

• machine learning
• classification
• feature selection
• rfe
• ann
• adaboost
• liver