LoRaWAN-Enabled Smart Campus: The Data Set and a People Counter Use Case

Smart Campus is one of the essential use cases in the Internet of Things (acrshort IoT). This work describes a long-range wide area network (LoRaWAN)-based smart campus data set comprising measurements of several sensors in hundreds of acrshort IoT devices. In addition, the data set contains information about the PHY and MAC layers of the LoRaWAN network. Therefore, we first describe the LoRa network, the connection between devices and gateway, and the gateway and network server. As the wireless network is prone to errors, e.g., outages, collisions, and interference, among other factors, the collected data may contain missing transmissions. To alleviate the problem of missing values, we resort to a
$k$
-nearest neighbor approach. For example, once the data imputation phase is complete, we employ an long short-term memory (LSTM) architecture to predict future sensor readings. Then, we build a deep neural network (DNN) to predict the room occupancy based on the selected sensor’s readings. Our results show that our model achieves an accuracy of 95% in predicting the number of people in a room. Furthermore, the data set is openly available and described in detail, which is an opportunity to explore other features and applications in Smart Campus.