An explainable multi-objective hybrid machine learning model for reducing heart failure mortality.

Area of Science:

• Cardiology
• Artificial Intelligence
• Machine Learning

Background:

• Heart failure (HF) is a leading global cause of mortality, necessitating early diagnosis for improved survival.
• Current machine learning (ML) and feature selection methods struggle with novel data and complex patterns.

Purpose of the Study:

• To introduce an advanced Multi-objective Stacked Enable Hybrid Model (MO-SEHM) for enhanced early heart failure detection.
• To improve the accuracy and robustness of ML models in diagnosing heart failure using feature selection.

Main Methods:

• Developed a MO-SEHM integrating a Stacked Enable Hybrid Model (SEHM) classifier with Non-dominated Sorting Genetic Algorithm II (NSGA-II) for multi-objective feature selection.
• Evaluated six ML models, including SEHM with and without NSGA-II, on a heart failure dataset from Faisalabad Institute of Cardiology (FIOC).
• Utilized Local Interpretable Model-agnostic Explanations (LIME) to ensure model transparency.

Main Results:

• The MO-SEHM demonstrated superior performance compared to other models, achieving 94.87% accuracy.
• The model identified nine relevant features crucial for accurate heart failure diagnosis.
• The Pareto front analysis confirmed the effectiveness of the proposed MO-SEHM.

Conclusions:

• The MO-SEHM offers a significant advancement in early heart failure diagnosis through optimized feature selection and robust classification.
• The model's interpretability using LIME enhances trust and understanding for patients and stakeholders.
• This approach holds promise for improving patient outcomes in cardiovascular disease management.