This paper presents several room occupancy detection methods using statistical learning model. Occupancy detection system is mainly used for energy saving in green buildings such as offices and residential apartments. The system will automatically switched-off the lighting, heating or ventilation appliances when the room is empty. The proposed work uses temperature and humidity sensor to detect human presence. All the input values from this sensor are transmitted to an IoT platform called Blynk (for data monitoring), through the medium of an open-source microcontroller board NodeMCU. The collected data is analyzed using two different approaches which are supervised learning model and unsupervised learning model. Results show that for supervised learning, SVM performs slightly better than Decision Tree. While for unsupervised subspace learning, Minimax yields better probability of detection than SVD in worst case criterion.