A Machine Learning Based Study to Predict Depression with Monitoring Actigraph Watch Data

The consequences of depression are breathtaking these days. The suicidal tendency, as well as other fatigues, depression has almost soaked the world. A detection system can combat such consequences early. Motor activity sensor values carry out an individual's daily routine activities that can somewhat signify momentary changes in behavior. A consolidation of these motor sensor data with other demographic, clinical data can be very convenient in terms of depression detection. The combination of motor sensor reads as well as demographic data has been obligated in this study with machine learning approaches, namely Random Forest(RF), AdaBoost, and Artificial Neural Networks (ANN), achieving accuracy and Fl-score of 98% in both cases. The Cohen's kappa coefficient and Matthew's correlation coefficient are 0.96 in both factors.