IoT Aware Smart Agriculture Using Extreme Learning Classifier Based Predictive Analytics in a Cloud Environment

Abstract

Agriculture encompasses soil nurturing, crop cultivation, and influencing human life and the environment for global economic growth. Effective irrigation and soil moisture management directly impact crop yields. To optimize productivity, an IoT-based soil monitoring system analyses soil parameters and weather conditions, generating substantial data stored on cloud platforms for predictive analytics. However, traditional methods face challenges in accurate and timely predictions. Addressing this, a novel Proximity Scaling Laplace Kernelized Extreme Learning Classifier (PSLKELC) is proposed. It consists of three stages: preprocessing, dimensionality reduction, and classification. Preprocessing involves data cleaning and transformation. Dimensionality reduction utilizes McNemar statistic multidimensional scaling to select relevant variables. Finally, the Laplace kernelized Extreme Learning classifier predicts soil moisture using the reduced dataset. The experimental evaluation compares the PSLKELC method with conventional techniques, considering metrics like accuracy, mean absolute error, time, and space complexity across various data sample sizes. Results demonstrate that PSLKELC enhances soil moisture prediction accuracy with reduced time and space complexity compared to traditional methods.