Comparative Analysis Of Machine Learning Algorithms To Predict Liver Disease

Liver disease is a significant global health concern necessitating early detection and accurate diagnosis. This study focuses on the development and evaluation of a machine learning model for predicting liver disease based on patient data. The dataset comprises 583 pre-processed records, encompassing demographic information, laboratory results, and medical history. Among these, 416 records are classified as having liver disease and 167 as not having it. Machine learning algorithms, including logistic regression, decision tree, support vector machines, and K-nearest neighbours, are assessed using a 70-30 training-testing split. Performance metrics encompass Jaccard Score and F1 Score. Logistic regression attains the highest accuracy at 87.69%, with Jaccard and F1 Scores of 88.54% and 87.89% respectively. Decision tree and K-NN models achieve respectable accuracies of 81.54% and 80.77% correspondingly. Feature importance analysis identifies total bilirubin, direct bilirubin, and alkaline phosphatase levels as pivotal predictors. These results demonstrate the potential of machine learning for accurate liver disease prediction, aiding in early intervention and improved patient care. Further exploration of advanced techniques and ethical considerations are suggested for future work.