Efficiency measure of Machine Learning Algorithms on Liver Disease Diagnosis

Abstract

This The death rate in India is high due to Liver disease as a result of bad lifestyle, storage food, uncontrolled blood sugar, obesity, smoking, and consumption of alcohol and inhale of harmful gases. Earlier detection can reduce death rates and it also helps the doctors to give the proper treatment to the patients. The liver disease datasets are analyzed by using Machine learning algorithms for the accurate disease diagnosis. The datasets were collected and annotated from Visakhapatnam, Vijayawada and Tirupathi based on the major geographical regions of Andhra Pradesh that are North Coastal Andhra Pradesh, Central Andhra Pradesh and Rayalaseema respectively. Three datasets are named Visakhapatnam dataset, Vijayawada dataset and Tirupathi dataset based on geographical region. Visakhapatnam dataset contains 12 attributes and has 499 samples. Vijayawada dataset contains 12 attributes and has 600 samples. The Tirupathi dataset contains 7 attributes and has 243 samples. The selected Classification Algorithms that are Naive Bayes, Decision Tree, Random Forest, Support Vector Machines and Multi-Layer Perceptron are castoff for scrutinizing their efficacy based on Accuracy, Precision, Sensitivity, Specificity, F-Measure, ROC-Area, FPR, MAE, RMSE, RRSE, Kappa Statistic and Building Time in classifying liver patient's dataset. Classification performance is very high in the Decision Tree classification algorithm for Visakhapatnam and Tirupathi datasets, whereas Classification performance is very high in the Random Forest classification algorithm for the Vijayawada dataset. Building time is more for MLP in the Vijayawada dataset. This study motivated for the development of the Liver Diagnosis App using the Decision tree algorithm.