Improved Stress Classification Using Automatic Feature Selection from Heart Rate and Respiratory Rate Time Signals

Time-series features are the characteristics of data periodically collected over time. The calculation time-series helps in understanding underlying patterns and structure data, as well visualizing data. manual selection feature from a large temporal dataset time-consuming. It requires researchers to consider several signal-processing algorithms analysis methods identify extract meaningful given These core machine learning-based predictive model designed describe informative signal. For accurate stress monitoring, it is essential that these not only but also well-distinguishable interpretable by classification models. Recently, lot work has been carried out on automating extraction times-series features. In this paper, correlation-based algorithm proposed evaluated stress-predict dataset. calculates list 1578 heart rate respiratory signals (combined) using tsfresh library. then shortlisted more specific Principal Component Analysis (PCA) Pearson, Kendall, Spearman correlation ranking techniques. A comparative study conventional statistical (like, mean, standard deviation, median, mean absolute deviation) versus selected performed linear (logistic regression), ensemble (random forest), clustering (k-nearest neighbours) achieved higher performance with an accuracy 98.6% compared feature’s 67.4%. outcome suggests vital have better analytical rather than for classification.