Forecasting occupancy count using Recurrent neural network (LSTM)

The Internet of Things has advanced in such a way that smart homes become more of a reality. One of the advancements that make this possible is the use of efficient and cost-effective sensors that collect environmental data that can be used for machine learning. Environmental sensors are non-intrusive, so they can be deployed to homes, offices, and other similar environments with minimum risk of privacy. Recurrent learning models such as long short-term memory (LSTM) can be used to predict future occupancy information using environmental data for smart environment and security with high accuracy. This trained model can be used in other applications related to building occupancy. In this work, we trained a long short-term memory model using a publicly available dataset to forecast the occupancy count and achieved a high accuracy of 97.3%. We present the training process, the hyperparameters, and the results. This model can be used in other building occupancy estimation applications.