Diagnosis of Liver Disease Using ANN and ML Algorithms with Hyperparameter Tuning

Liver disease has caused a serious concern in India and is rising similar to diabetes, about 3.17% of Indians die due to this disease. It is always a tedious task to identify liver disease in the initial stages, which may lead to life-threatening diseases such as liver cirrhosis, cancer, etc. The main reasons behind it are alcoholism, obesity, hepatitis B and C, etc. Therefore, there is a need for enabling automatic detection in the early stages which in turn leads to diminishing the death rates. This study proposes novel machine learning and neural network models to unfold this problem. Here, the Indian Liver Patient Dataset, a UCI repository with 583 samples is used. Artificial Neural Network (ANN)and Machine Learning (ML)models such as K-Nearest Neighbor, Random Forest Classifier, Logistic Regression and Support Vector Machine are applied along with hyperparameter tuning to get the best model for accurate liver disease prediction. F1-score, accuracy, recall and precision with and without feature selection are analyzed. In this study, the F1-score of LR is the highest at 83.85% when no feature selection is applied. ANN and SVM topped 83.38% with the F1-score exhibiting the highest accuracy of 73.57%.