An Efficient Hybrid Data Mining Model for Prognostication of an Imbalanced Data Set of Liver Disorder: A K-Prototype Naïve Bayes Approach

Background: Liver disorders have recently become the deadliest disorder in many countries, with the number of patients increasing as a result of alcohol consumption, exposure to toxic gases, and ingestion of tainted foods and drugs. Data mining is the most effective approach for detecting the disease early on.Objective: This study aimed to predict and diagnose early-stage liver disorders.Method: In this study, we used the Indian liver patient dataset from the UCI machine learning repository. This dataset contains the sex imbalance for which we applied both oversampling and undersampling strategies; we used principal component analysis (PCA) for feature selection. In this research, we built eight models from 4 experiments in RStudio with the required packages. These models are compared based on the performance factors, which include accuracy, sensitivity, specificity, and error rate. We constructed the Naïve Bayes model and a new innovative hybrid model combining k-prototype clustering and the Naïve Bayes classifier (K-PNB).Results: The hybrid model gave a classification accuracy of 94%, a sensitivity of 99%, a specificity of 90% and a low error rate of 0.05%.Conclusion: The findings showed that the proposed hybrid model (the K-PNB) outperformed the other models, which detect and diagnose liver disease in the early stages in very little time. How to cite this article:Divya, Singh V, Dohare R, Kumar M. An Efficient Hybrid Data Mining Model for Prognostication of an Imbalanced Data Set of Liver Disorder: A K-Prototype Naïve Bayes Approach. Chettinad Health City Med J. 2024;13(4):21-33. DOI: https://doi.org/10.24321/2278.2044.202456