Feature Predominance Ensemble Inquisition towards Liver Disease Prediction using Machine Learning

Authors:
  • VCVamsi Chowdary
    Vamsi Chowdary
  • SSSagar Srivastava
    Sagar Srivastava
Preprint
Publisher website
Google Scholar
Abstract
The food digestion is very important criteria for the organs to coordinate the body activity. The toxic substances are removed from the body organs with the proper digestion and its enzymes. The functionality of the liver is directly bonded with the digestion ability of the body. With this overview, this paper attempts to analyse each clinical features that influence the target with the following contributions. Firstly, the liver dataset from UCI machine repository is subjected with the data processing. Secondly, the exploratory data analysis is done by finding the correlation of all the features in the dataset. Thirdly, the data is applied with Gradient Boost Regressor to find top feature importance and then fitted to all the classifiers to analyze the performance metrics before and after feature scaling. Fourth, the data is applied with Extra Tree Regressor to find top feature importance and then fitted to all the classifiers to analyze the performance metrics before and after feature scaling. Fifth, the data is applied with Random Forest Regressor to find top feature importance and then fitted to all the classifiers to analyze the performance metrics before and after feature scaling. The implementation is done in python scripts with Anaconda Spyder IDE. Experimental results shows that Random Forest Classifier is found to retain the accuracy of 74% before and after feature scaling with the top features from Gradient Boost Regressor. SGD Classifier is found to raise the accuracy of 72% after feature scaling with the top features from Extra Trees Regressor. Bagging Classifier is found to retain the accuracy of 72% before and after feature scaling with the top features from Random Forest Regressor.