An Improved Machine Learning Model for Stress Categorization

Stress and depression are now a global problem. 25% of world's population is facing this problem. With the growing use of sensor equipped intelligent wearables, physiological parameters can be easily extracted and effectively analyzed to predict it at an early stage. Stress management is a complex problem, as to predict stress; there exist several parameters to be considered. Choosing the right parameter is a challenging task, to predict the stress more accurately. In this work, to select the most efficient parameters the unsupervised algorithm, K-Means is used; and after getting the right parameters, a radial basis function based neural network is utilized to group the captured data to be stressed or non-stressed. The model also identifies the type of the stress. The work is validated in Python based environment and gives a promising result, in terms of accuracy.