Machine Learning Methods and Their Applications in Biomedical Sensors
TL;DR
Research indicates that RFR possesses strong anti-overfitting capabilities and it’s suitable for malaria detection, gait disorder identification, and other scenarios; PLS efficiently processes high-dimensional collinear data and performs well in hemoglobin testing and cfDNA analysis; PR adapts to nonlinear relationships and offers fast computation, meeting the needs of tuberculosis and dopamine detection.
Abstract
. In recent years, biomedical sensors have achieved significant breakthroughs. Machine learning technology has facilitated advancements in sensor applications by improving signal quality and automatically extracting features. This article systematically reviews relevant research, focusing on three machine learning methods, including Random Forest Regression (RFR), Partial Least Squares (PLS), and Polynomial Regression (PR), and provides a comprehensive comparison in terms of performance, advantages, disadvantages, and application scenarios. Research indicates that RFR possesses strong anti-overfitting capabilities and it’s suitable for malaria detection, gait disorder identification, and other scenarios; PLS efficiently processes high-dimensional collinear data and performs well in hemoglobin testing and cfDNA analysis; PR adapts to nonlinear relationships and offers fast computation, meeting the needs of tuberculosis and dopamine detection. This study offers insights for researchers. In the future, with developments of flexible integration, multi-modal fusion, and non-invasive continuous monitoring technology, machine learning methods will be further refined to accommodate higher-quality sensing data, leading to broader application of biomedical sensors.