HIERARCHICAL DISCRIMINATIVE AI FOR EFFICIENT CROP RECOMMENDATION AND ON-DEMAND FORECASTING IN PRECISION AGRICULTURE

Abstract

Agriculture plays a pivotal role in country economy growth as it is major resource of food, employment and raw materials. Crop recommendations play an essential role in precision agriculture to predict the most appropriate crops. However, numerous factors such as soil characteristics, climate conditions, make it difficult to accurately predicting suitable crop. In this paper, Polynomial REgressive Momentum Optimized Discriminative AI (PREMOD-AI) model is introduced for accurately predicting crop. During acquisition phase, data samples are collected. Hierarchical Discriminative AI model includes three kinds of layers such as input layer, three hidden layers and output layer. First, collected data are given to input layer. Subsequently, data preprocessing is performedin hidden layer 1. Dimensionality reduction process is completed at hidden layer 2 by applying Hockey-stick indexed regression to determine most relevant features. Classificationis performed using Chatterjee correlation coefficients in hidden layer 3. The fine tuning process is employed to minimize error using horse herd optimization method. Finally, accurate croppredictions are obtainedat output layer. Experimentis conducted with accuracy, precision, recall, F-score, RMSE, MCC, R2score and crop recommendation time. The result of PREMOD-AI model demonstrates an improvement in accuracy by 95%, precision by 95%, and F-score by 96%, along with a 97% increase in recall. In addition, it reduces crop recommendation time by 76ms, RMSE by 0.053, MCC by 0.91 and R2score by 0.92 when compared to traditional deep learning approaches.