Stress Detection using Machine Learning and Deep Learning

Stress is a normal phenomenon in today's world, and it causes people to respond to a variety of factors, resulting in physiological and behavioural changes. If we keep stress in our minds for too long, it will have an effect on our bodies. Many health conditions associated with stress can be avoided if stress is detected sooner. When a person is stressed, a pattern can be detected using various bio-signals such as thermal, electrical, impedance, acoustic, optical, and so on, and stress levels can be identified using these bio-signals. This paper uses a dataset that was obtained using an Internet of Things (IOT) sensor, which led to the collection of information about a real-life situation involving a person's mental health. To obtain a pattern for stress detection, data from sensors such as the Galvanic Skin Response Sensor (GSR) and the Electrocardiogram (ECG) were collected. The dataset will then be categorised using Multilayer Perceptron (MLP), Decision Tree (DT), K-Nearest Neighbour (KNN), Support Vector Machine (SVM), and Deep Learning algorithms (DL). Accuracy, precision, recall, and F1-Score are used to assess the data's performance. Finally, Decision Tree (DT) had the best performance where DT have accuracy 95%, precision 96%, recall 96% and F1-score 96% among all machine learning classifiers.