Machine Learning based Data Traffic Prediction and Classification of Sensor Nodes in Wireless Sensor Network

The lifetime and high throughput are two significant concerns of wireless sensor networks. The lifetime of the network can be increased by the efficient allotment of the slots and efficient utilization by sensor nodes. However, the major challenge is that the data generation by the sensor nodes is primarily random in real-time applications. Therefore, the allotted channel is not utilized efficiently, and most of the time, the channel’s bandwidth is wasted due to a variable rate of data generation. In such cases, few nodes fail to transmit the data, and few nodes waste their slots due to the unavailability of the data. To resolve the issue, a machine learning-based efficient data slot allotment method is proposed for energy-efficient bandwidth utilization in the present work. In the proposed method, the node’s data traffic is predicted based on the previous dataset records. In addition, the prediction helps classify the nodes into event monitoring or periodic/continuous monitoring nodes. The dominance of the proposed protocol is presented by the energy consumption of the presented protocol compared to the energy-efficient bit-map-assisted protocol. The results prove that the proposed protocol saves energy by efficiently allotting data slots based on the machine learning approach.