Forecasting indoor temperature for smart buildings with ARIMA, SARIMAX, and LSTM: A fusion approach

The surging demand for power usage in the last two decades has increased exponentially, mostly in the building sector, due to people's high standard of living. The energy usage of a building is dependent on various surrounding parameters like temperature, humidity, appliance usage, and many more. Temperature forecasting inside smart building premises may reduce energy consumption as well as other associated factors. In our current research work, we have used different models based on internet of things (IoT) data for the estimation of indoor temperature. The data for this work has been collected from temperature sensors deployed inside smart buildings. For prediction, we have used various models (ARIMA, SARIMAX, and LSTM). Applying a single model along with collected data may not be efficient for accurate prediction. Therefore, the present research proposes a FUSION approach, i.e., a combination of (ARIMA, SARIMAX, and LSTM) for more accurate temperature prediction. The evaluation criteria of the proposed model are based on the MAPE (Mean Absolute Percentage Error) and MSE (Mean Squared Error) metrics. The values are compared for each individual and the proposed model to get the lowest error rate of prediction. Even though the results of other models showed a good forecast, the FUSION approach did much better than other models, which shows how well this research was done.