A CRISP-DM Approach for Predicting Liver Failure Cases: An Indian Case Study

Abstract

We are currently living a period in which data processing and analysis are increasingly relevant and the health sector is no exception. In this way, through data mining processes, it is possible to make a number of predictions in the medical field, such as predicting medical conditions and disease progression. Acute Liver Failure (ALF) is a rare but critical disorder associated with high mortality. The aim of this work is to predict cases of acute hepatic insufficiency based on clinical data through data mining techniques. To this end, the CRISP-DM methodology was followed, in which five classifiers were applied, namely, Decision Tree, k-Nearest Neighbor, Random Forest, Rule Induction, and Naïve Bayes. Throughout this work, the RapidMiner software was used and the different models were analyzed based on Accuracy, Precision, Recall, Kappa Statistic, and Specificity. The best data mining model achieved an Accuracy of 0.925, a Precision of 0.869, a Recall of 1.000, a Kappa of 0.849, and a Specificity of 0.849, using split validation and the k-Nearest Neighbor algorithm.