Identification of Specific Human Behaviours from Cortisol profiles using a Neural Network Based Ensemble Classifier

Many vital bodily functions like stress response, immune response, blood pressure monitoring, etc. are controlled by the hormone cortisol. Variations in this hormone are related to various disorders physiological and psychological disorders. Therefore, there is a need to understand the impact of daily activities on this hormone. In this work, we designed a neural network-based classifier ensemble model that attempts to identify if the subject engaged in a particular activity in the past 48 hours, when provided with the cortisol levels measured in the same duration. The activities/behavior we focused on in this preliminary study is caffeine, nicotine and other (non calorie rich) food consumption. We were able to obtain f-score of 0.74, 0.71 and 0.68 respectively. The initial study results seem to demonstrate a predictive correlation between various daily habits and cortisol, and a possibility of isolating specific past activities that impacted the measured values of cortisol. This lays the foundation of exploring other such correlations in depth, that can lead to regulation of cortisol and other hormones.