Multi-view ensemble learning with empirical kernel for heart failure mortality prediction.

Area of Science:

• Cardiology
• Artificial Intelligence
• Machine Learning

Background:

• Heart failure (HF) presents a significant global health challenge with high prevalence.
• Accurate mortality prediction in HF patients is crucial for effective clinical management.
• Existing prediction models may not fully leverage complex patient data.

Purpose of the Study:

• To develop and validate a novel multi-view ensemble learning algorithm (MVE-EK) for predicting heart failure patient mortality.
• To enhance the accuracy and reliability of mortality prediction using hospital records.
• To provide clinicians with data-driven insights for personalized HF treatment strategies.

Main Methods:

• Proposed MVE-EK algorithm utilizes multi-view ensemble learning and empirical kernel mapping.
• Patient features are divided into multiple views, and samples are under-sampled into balanced subsets.
• Empirical kernel mapping transforms data into a linearly separable kernel space for improved classification.
• Ensemble learning is performed using a designed inter-view loss function.

Main Results:

• The MVE-EK algorithm demonstrated superior performance compared to existing algorithms on three real-world HF datasets.
• Validation was conducted on 10,203 hospitalization records from 4,682 HF patients (March 2009 - April 2016).
• The algorithm effectively reduced data imbalance and improved prediction accuracy.

Conclusions:

• The MVE-EK algorithm offers a robust and effective method for predicting heart failure patient mortality.
• This predictive capability can significantly assist clinicians in developing personalized treatment plans.
• The study highlights the potential of advanced machine learning techniques in cardiovascular disease management.