Prediction of Human Performance from Brain Signal Driven by Different Indoor Room Temperatures

It is important to understand how indoor environments influence office worker's productivity and develop a room temperature control system to improve human productivity. It will have a significant impact on our society. In this thesis, we designed a simulated office-work experiment and attempt to predict human productivity under different room temperature (22 °C and 30 °C). Seven healthy adult participants were recruited. During the experiment, EEG, skin temperature, heartbeat rate and thermal survey questionnaire were recorded and compared. By using R2 to compare each factors' correlation, we decide to investigate the effectiveness of EEG signals to predict subjects' simulated office-work tasks with LASSO, SVM and neural network. 10 folder cross validation is used to evaluate models. With EEG frequency features, we use two different performance indexes and obtain around 80% and 75% accuracy by LASSO, 77% and 74% accuracy by SVM, 87% and 85% accuracy by neural network model separately. Our work demonstrates the potential of accurately predicting office worker's productivity by using EEG signals and neural network model.