Liver Disease Detection and Prediction Using SVM Techniques

Because of the high prevalence and mortality rate of liver disease, early diagnosis is essential. Machine learning techniques can be introduced in the diagnosis and treatment of diseases. In this paper, several machine learning techniques Random Forest, SVM, Logistic Regression, AdaBoost, Decision Tree, Gradient Boosting, and K Neighbors are used for predicting liver disease. The Liver Patient Data Set (ILPD) is used to test these algorithms. The preprocessing procedures conducted on a dataset in this work are data cleaning, normalization, removing outliers, imbalanced data processing using SMOTE method, and important feature selection using a random forest. The results show that the algorithms decision tree, AdaBoost, and random forest have the highest accuracy of 100% during the training phase. However, during testing, the algorithm random forest has the highest accuracy of 80.36%. It is noticeable also from results obtained in this paper that some algorithms such as SVC and logistic regression overcome the problems of overfitting and underfitting but other algorithms including decision tree, Adaboost, and gradient boost have slight problems in overfitting and underfitting.