Prediction of Liver Disease Using Machine Learning Approaches Based on KNN Model

The liver, the most crucial interior organ of the human body, performs functions of metabolism control and food digestion. Liver disorders can sometimes prove fatal, and appropriate treatment at the right time will save many lives. Research has been conducted to predict and diagnose liver diseases for quite a while, and Machine Learning (ML) tools have proven very effective. We have considered eight ML models for this study while working on the Liver Patient Dataset, which contains more than 30k instances, for accurately predicting liver diseases. This study includes some boosting algorithms as well. For the proper judgment of the performance of the proposed models, commonly used performance metrics such as accuracy, RoC-AuC, F1 score, precision, and recall have been used. We have inferred that the k-Nearest Neighbor (KNN) produced the most accurate results at 92.345%. Since the models are not overfitted, they are k-fold cross-validated, and hence, the standard deviation of each model is lower. Therefore, the standard deviation of KNN is 0.815, and the False Negative (FN) rate comes out to be 3.7%.