Implementation of Intelligent Deep Learning Approach for Multiple Class Stress Identification by using Observed Variable Heart Signals

Abstract

Stress is a one of the crucial physiological and psychological features that shows 
more impact on health conditions of person and also shows impact on life style. 
Most of the conventional stress evaluation approaches are depends on 
individual matters, fail to monitor continuous data samples and deviated the 
predictions, mostly these approaches focused on binary classification and not 
supporting multiple classification. In this research we Implementation of 
Intelligent Deep Learning Approach for Multiple Class Stress Identification by 
using Observed Variable Heart Signals (IDLMCSI) approach, Variability in 
heart rate is gathered in the form of HRV signals through IoT wearable devices, 
dynamic data samples are processed using feature engineering approaches and 
then apply 1D CNN to learn features that belonging to stress, identified various 
kinds of stress movements and classified into various types classes. 
Performance of proposed IDLMCSI model is evaluated and then compared with 
Hybrid CNN–LSTM, LSTM, GRU, ANN, RF, SVM, DT, KNN, and NB. HRV 
signal patterns are not captured by conventional ML approaches, proposed 
IDLMCSI exhibits high percent of accuracy, high percent of recall, high percent 
of F1-score, and high percent of precision due to optimization of functions and 
as well as supporting convolutional FL approach, proposed IDLMCSI exhibits 
low value of log loss and that represents more accurately predict classes with 
high confidence value, exhibits high value of Cohen’s Kappa and exhibits high 
value of MCC due to optimization of functions and as well as supporting 
convolutional FL approach.