Umematsu, T., Sano, A., Taylor, S., and Picard, R. "Improving Stress Forecasting using LSTM Neural Networks." The 40th International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Honolulu, Hawaii, July 2018.

Abstract

Accurately forecasting stress might enable people to make changes that could improve their future health. In this paper, we examine how accurately previous N-day multi-modal data from wearable sensors, mobile phones and surveys can predict tomorrow’s level of stress using long short-term memory neural network models (LSTM), logistic regression (LR), and support vector machine (SVM). Using 1231 days of data from 201 college students, we find the LSTM outperforms the LR and SVM with the best results reaching 83.3-84.1% using 5-7 days of prior data.