An Intelligent Model Based on Data Mining and Machine Learning Approaches for Diagnosing Hepatitis Disease

Hepatitis disease can be fatal if it is not diagnosed in its early stage. Hence, this study introduces an intelligent model to predict hepatitis in early stage and to do so applies an adaptive Multi-layer Neural Network (MLNN) as classifier along with missing value imputation technique and well-known feature selection and reduction methods. Fuzzy K-means (FK-means) method has been employed for missing value imputation followed by multiple feature selection and reduction methods such as Recursive Feature Elimination (RFE), Chi-Square test, Least Absolute Shrinkage and Selection Operator (LASSO), Principal Component Analysis (PCA) and Genetic Algorithm (GA). This study uses three benchmark datasets from UCI Machine Learning Repository to verify the proposed model. Most of the studies have provided their model performance using only accuracy but this is not sufficient to measure the performance of a model as it can be misleading for imbalanced dataset. For that reason, the performance of the proposed model is evaluated based on the accuracy, precision and recall. Performance analysis reveals that GA-MLNN provides good results for hepatitis dataset by achieving 90.32% accuracy, 89.66% precision and 100% recall. However, for other two datasets, Chi-Square-MLNN performs better than others.