An IoT Driven HTM Based Predictive Model for Early Detection of Sheath Blight Disease in Paddy Crops

Abstract

Rhizoctonia solani is a highly destructive fungal disease which can cause as much as 50% paddy loss. Timely and timely prevention and early prediction of these crop diseases are crucial to reduce losses and to foster sustainable agriculture. This study presents a new Internet of Things (IoT) based predictive model of sheath blight appearance using Hierarchical Temporal Memory (HTM), which mimics the human neocortex function to predict the probability of sheath blight occurrence in advance of the appearance of visible symptoms. Using IoT sensors, time-series information of the environment, such as temperature, humidity and rainfall in paddy fields were collected continuously. The HTM model was used to learn temporal patterns to predict disease occurrence using these parameters as input. The model was applied and tested with the data collected from the Cuddalore district of India from 2019 to 2023. The proposed HTM-IoT framework resulted in a consistent increase of the prediction accuracy as indicated, reaching the highest accuracy of 94% in 2023. Using correlation analysis, it was noted that there was a strong correlation between the environmental variables and the severity of the disease, such as humidity (0.86), rainfall (0.83) and temperature (0.79). The resulting model can be used to detect sheath blight outbreaks early, helping farmers to take timely actions to control it, minimize the use of pesticides, and move towards precision agriculture to ensure sustainable crop production.