Machine Learning Techniques for the Performance Enhancement of Multiple Classifiers in the Detection of Cardiovascular Disease from PPG Signals

This research explores the application of machine learning techniques to improve cardiovascular disease (CVD) detection using photoplethysmography (PPG) signals. Dimensionality reduction techniques, including Hilbert transform, nonlinear regression, and heuristic-based methods (ABC-PSO, cuckoo search, dragonfly optimization), were applied to reduce PPG data size. Twelve classifiers (PCA, EM, logistic regression, GMM, BLDC, firefly clusters, harmonic search, DFA, PAC Bayesian learning, KNN-PAC Bayesian, SDC, detrended SDC) were used to classify CVD and normal PPG segments. The Hilbert transform combined with the harmonic search classifier achieved the highest accuracy (98.31%) and good detection rate (96.55%).

Simon W Rabkin, Andrea Cataldo, Sivamani Palanisamy, Harikumar Rajaguru