A non-linear SVR-based cascade model for improving prediction accuracy of biomedical data analysis

Area of Science:

• Biomedical data analysis
• Machine learning applications in healthcare

Background:

• Biomedical data analysis is crucial for diagnosis, treatment, and monitoring.
• Current machine learning (ML) methods struggle with large datasets, requiring excessive resources or lacking sufficient accuracy.
• There is a need for efficient and accurate methods for analyzing large-scale biomedical data.

Purpose of the Study:

• To develop a novel ensemble model for accurate approximation of large biomedical datasets.
• To address the limitations of existing ML methods in terms of resource consumption and accuracy.
• To improve the efficiency and effectiveness of biomedical data analysis for applications like stress level determination.

Main Methods:

• Developed a new ensemble model based on cascading ML methods and response surface linearization.
• Utilized Ito decomposition for nonlinear input expansion at each model level.
• Employed Support Vector Regression (SVR) with a linear kernel as the weak learner.

Main Results:

• The developed SVR-based cascade model achieved over 20 times higher accuracy (based on Mean Squared Error - MSE) compared to existing methods.
• Demonstrated a significant reduction in the training procedure duration.
• Successfully applied the model to predict heart rate from a large, real-world biomedical dataset.

Conclusions:

• The proposed SVR-based cascade model offers a highly accurate and efficient solution for analyzing large biomedical datasets.
• The model's ability to predict heart rate aids in determining human stress levels, with broad applied potential.
• This approach overcomes the resource and accuracy limitations of traditional ML methods for complex biomedical data analysis.