Proximal Weighted Correlative Sequential Extreme Learning Machine For Iot-Based Automatic Crop Prediction In Agriculture

Abstract

Agriculture is rehearsal of cultivating soil, rising crops, and hoisting animals for food, fiber, other products employed to maintain as well as improve human life. It is a key development in India’s economy and social fabric, continuously evolving to meet the demands of a growing population and changing environmental conditions. Smart agriculture, also known as precision agriculture, leverages advanced technologies to enhance crop productivity by making farming practices more efficient, sustainable, and data-driven. Through utilizing tools namely IoT as well as information analytics, farmers observe and control their fields through unprecedented precision. Therefore, analyzing the soil and environmental parameters is crucial for optimizing crop forecast. Different ML methods have been designed for crop forecast, however the major challenging issues faced by existing techniques are accuracy levels, error, and complexity. In order to solve these existing issues, a novel proximal weighted correlative sequential extreme learning (PWCSEL) model is developed. Major aim of PWCSEL method is to improve accuracy of crop prediction with minimal error and complexity. The proposed PWCSEL model involves four major steps namely data acquisition, pre-processing, feature selection, classification. In data acquisition stage, IoT involves using interconnected tools as well as sensors to gather soil and environmental characteristics from the farming environment. These IoT-acquired data are stored in a dataset repository for further processing. Next, data pre-processing is carried out to handle missing data and outliers within the dataset. After pre-processing, the significant feature selection process is employed in PWCSEL model to reduce time and space complexity during the prediction process. This selection involves identifying the most pertinent features that contribute to the predictive accuracy using Spatially Uniform Rosenthal weighted Correlative Relief algorithm. With the selected pertinent features, classification is performed using the Hamann indexive sequential extreme learning model for accurate crop prediction with minimal error. The observed results reveal PWCSEL increases the prediction accuracy and substantial reduction in error rate and prediction time compared to conventional methods.