ESVM-SWRF: Ensemble SVM-based sample weighted random forests for liver disease classification

Area of Science:

• Medical data mining
• Machine learning in healthcare
• Disease prediction and diagnosis

Background:

• Healthcare faces challenges with large datasets for disease analysis and prediction.
• Existing data mining models have limitations including high execution time, computational complexity, and slow convergence.
• Effective data transformation into valuable insights is crucial for accurate medical decision-making.

Purpose of the Study:

• To propose an advanced ensemble model for predicting liver diseases.
• To address the shortcomings of existing methods in terms of efficiency and accuracy.
• To optimize model parameters using a novel metaheuristic algorithm.

Main Methods:

• Proposed an ensemble SVM-based sample weighted random forests (eSVM-swRF) model.
• Utilized extraction, loading, transformation, and analysis (ELTA) for data pre-processing.
• Employed a novel improved colliding body optimization (NICBO) algorithm to optimize eSVM-swRF parameters (P, T, mTry).
• Validated the model using the UCI liver disease dataset and RapidMiner Studio.

Main Results:

• The eSVM-swRF model with NICBO optimization demonstrated outstanding performance.
• Achieved superior prediction accuracy compared to existing methods like PSO-SVM, FuzzyANWKNN, NB-SVM, and Neural Network.
• The proposed method effectively handles complex medical data for disease prediction.

Conclusions:

• The developed eSVM-swRF with NICBO algorithm offers a highly effective solution for liver disease prediction.
• This approach overcomes limitations of previous methods, providing a scalable and efficient diagnostic tool.
• The study highlights the potential of advanced machine learning techniques in medical data mining for improved healthcare outcomes.