Understanding the suitability of parametric modeling techniques in detecting the changes in the HRV signals acquired from cannabis consuming and nonconsuming Indian paddy-field workers

The current study aims to recognize the suitability of the parametric modeling methods in identifying the possible alterations of the heart rate variability (HRV) signals induced by regular intake of cannabis (bhang). The 5-s HRV signals were extracted from the electrocardiogram (ECG) data of 200 paddy-field workers. The paddy-field workers comprised 100 cannabis consumers and 100 control volunteers. The parametric models of the HRV signals were developed employing autoregressive (AR), moving average (MA), and autoregressive moving average (ARMA) methods. None of the model parameters were significantly different among both groups, as analyzed by Mann–Whitney U test and t-test. After that, the weight-based feature ranking and dimension reduction methods were employed for selecting suitable parameters for developing machine learning (ML) classifiers. Based on the scrutiny of the performance metrics, a Logistic Regression (LR) classifier that employed AR model parameters generated the best classification efficiency. Independent Component Analysis (ICA) was used for feature selection in the development of the LR classifier.