A lightweight deep learning and whale optimization framework for sustainable precision agriculture

Abstract

The changing needs of the modern agriculture require smart and resource saving solutions to such problems as falling productivity, irresponsible use of inputs, and deterioration of the environment. This paper presents a hybrid framework AgriCLWO-Net, which consists of lightweight Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM) model along with Whale Optimization Algorithm (WOA) to provide precision agriculture services by sensors integrated Internet of Things (IoT) environments. The suggested model will categorize the health status of crops, optimize irrigation and spreading of fertilizers, and enhance sustainability performance based on spatiotemporal field information. The methodology takes advantage of CNN to perform spatial patterns area recognition based on multisensory stimuli, LSTM to perform temporal relationships in crop and atmospheric patterns, and WOA to tune the hyperparameters and adaptive decision-making. The model was tested against a sample dataset of the Indian agricultural areas including the temperature, soil moisture, humidity, and nutrient measurements. Findings show that the classification accuracy (98.54%), water use efficiency (27.93%), fertilizer reduction (21.64%), and sustainability index increased (0.54 to 0.76) significantly as compared to the existing baseline models.