Prediction of different types of liver diseases using rule based classification model

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Data Mining for Disease Prediction

Background:

• Clinical diagnosis of liver diseases involves numerous complex laboratory tests.
• A simplified, accurate prediction model is needed to aid in classifying liver diseases.
• This study introduces a Rule Base Classification Model (RBCM) integrating rules and data mining.

Purpose of the Study:

• To propose a novel rule-based classification model for predicting liver diseases.
• To leverage machine learning techniques for enhanced liver disease diagnosis.
• To compare the efficacy of various data mining algorithms within the proposed model.

Main Methods:

• A dataset of 583 patients (441 male, 142 female) with 12 attributes was utilized.
• Machine learning techniques including Support Vector Machine (SVM), Rule Induction (RI), Decision Tree (DT), Naive Bayes (NB), and Artificial Neural Network (ANN) were employed.
• K-cross fold validation and statistical tests (ANOVA, Chi-square) were used for performance evaluation and data analysis.

Main Results:

• The Decision Tree (DT) technique within the rule-based model achieved the highest accuracy (98.46%), sensitivity (95.7%), specificity (95.28%), and Kappa (0.983).
• Support Vector Machine (SVM) showed the poorest performance with 82.33% accuracy.
• The rule-based model significantly outperformed models without rules, demonstrating the value of integrated rule-based approaches.

Conclusions:

• The proposed rule-based classification model significantly improves liver disease prediction compared to models without rules.
• The Decision Tree (DT) technique integrated into the rule-based model offers the most accurate diagnostic results.
• This model serves as a valuable tool for supporting medical decision-making in liver disease diagnosis.