Classification of Quality of Granary using Machine Learning based on Software-Defined Wireless Sensor Network

In India, several million tons of grain are stored in silos, warehouses, and gunny bags for future use. Temperature and grain moisture content are the most important factors responsible for the growth of mold, fungus, insects within the stored grain. If these factors are not monitored at early stages, contamination and insect infestation will damage all stored grain. In addition to these factors, CO2 evolution is also taken as a measure to detect deterioration in grain. The sensor network is a prominent technique through which environmental factors can be monitored. Although, this technique faces issues such as constraints of devices like limited energy, computational capability, and data storage. To overcome such issues, this paper proposes an architecture and methodology for monitoring temperature, moisture content, and CO2 concentration for granary called as Software-Defined Wireless Sensor Network. Data aggregation and data smoothing algorithms has proposed to reduce energy consumption by a node and to reduce noise. To predict the quality of stored grain, machine learning algorithms such as K-Nearest Neighbor, Random Forest, and Linear Regression has implemented at controller in control plane. Results show that Random Forest performs better and correctly predicted the quality of grain with the highest accuracy among all classifiers.