Optimizing ensemble machine learning models for accurate liver disease prediction in healthcare

Area of Science:

• Medical Informatics
• Computational Biology
• Machine Learning

Background:

• Liver disease, encompassing conditions like hepatitis and cirrhosis, often progresses silently, making early detection crucial.
• Risk factors include infections, alcohol abuse, obesity, and genetic predispositions.
• Effective management relies on early detection and lifestyle changes to prevent severe liver damage.

Purpose of the Study:

• To evaluate the efficacy of ensemble machine learning classifiers for predicting liver disease.
• To compare the performance of Random Forest, Ada Boost, and Gradient Boosting Classifiers using a specific dataset.
• To identify the optimal machine learning model for accurate liver disease diagnosis.

Main Methods:

• Feature extraction and selection were performed on the dataset.
• Hyperparameter tuning was conducted using Randomized Search CV and GridSearchCV.
• Ensemble classifiers including Random Forest, Ada Boost, and Gradient Boosting were implemented and compared.

Main Results:

• The Random Forest Classifier, optimized via GridSearchCV, achieved the highest accuracy exceeding 85.17%.
• This model demonstrated balanced performance with precision at 0.85, recall at 0.81 (presence) and 0.87 (absence), and F1-scores of 0.83 and 0.85.
• Ada Boost and Gradient Boosting Classifiers also showed high performance, though not significantly outperforming Random Forest.

Conclusions:

• Ensemble machine learning techniques show significant potential for diagnosing liver diseases.
• The optimized Random Forest Classifier can serve as a precise diagnostic tool, improving patient outcomes.
• Further research with larger datasets and deep learning integration is warranted to advance ML applications in healthcare diagnostics.