Design and Development of Efficient Ensemble Expert Systems for Crop Yield Prediction using Machine Learning Analytics

Abstract

Agricultural production is widely recognized as one of the largest occupations in India and plays a crucial role in India s economic growth. The increasing demand for sustainable agriculture and sustainable food availability in the face of climate variability and resource constraints has intensified the need for intelligent, data-driven decision support systems in agriculture. One of the most critical applications is Crop Yield Prediction (CYP), which aids in forecasting agricultural productivity prior to harvest. To ensure future food supplies, CYP provides the best decision-making to assist farmers in agricultural yield forecasting efficiently. Nevertheless, CYP is a difficult endeavour because of the intricacy of the underlying mechanisms and the effect of numerous factors, including weather patterns, soil characteristics, and crop management practices. However, agricultural datasets are inherently heterogeneous, noisy, and high-dimensional, with missing values, outliers, and regional dependencies that complicate predictive modeling yet essential for effective policy, resource management, and supply chain efficiency. To address these issues, this thesis presents a comprehensive research framework for the development of a robust, scalable, and interpretable ensemble expert system for CYP using Machine Learning (ML) paradigms, wherein Ensemble Learning (EL) serves as a specific approach. The research is structured across eight progressive chapters, each addressing specific challenges related to data quality, feature engineering, model design, and prediction accuracy. The study is grounded in real-world agricultural data from Assam, India, which includes meteorological, soil, and crop-specific variables of all Assam s districts (33 districts) over 15 years (2003-2017), making it both regionally relevant and practically significant.