Crop yield prediction using ensemble learning with effective data analytics

Purpose Accurate crop yield prediction (CYP) is essential for enhancing agricultural productivity, ensuring food security and enabling sustainable resource management. Machine learning (ML) algorithms have become popular in CYP because they can estimate crop production based on different characteristics, such as environmental characteristics, soil fertilizers and crop management. The integration of advanced data analytics with ensemble learning (EL) remains underexplored. This study aims to address this gap by demonstrating the effectiveness of EL-based approaches for reliable and accurate CYP, aiming to achieve a better outcome of the model. Design/methodology/approach This work introduces an artificial intelligence (AI) model named Precision Crop Yield Prediction AI System (PCYPAIS), which has been implemented to identify the most suitable crop for specific agricultural regions. It integrates an advanced preprocessing pipeline, including expectation–maximization (EM) algorithm to handle missing data, isolation forest (IF) technique to remove outliers, particle swarm optimization (PSO) technique for choosing relevant features and robust scaling (RS) technique for normalizing data, with an advanced EL technique named Extra Tree (ET) algorithm for classification. Findings The model is evaluated using a dataset from International Crops Research Institute for the Semi-Arid Tropics (ICRISAT) and benchmarked against multiple EL models, standard ML algorithms and state-of-the-art (SOTA) models. Results demonstrate that PCYPAIS consistently achieves superior prediction accuracy while addressing overfitting, variance and incomplete data challenges. These findings highlight the potential of integrating advanced data analytics with EL to enable data-driven decision-making, support sustainable crop planning and enhance agricultural productivity. Originality/value This work presents a novel EL-based CYP framework demonstrating the value of integrating data preprocessing, feature selection and ensemble-based classification. The technical significance of this work lies in showing that preprocessing is not merely an auxiliary step but a critical determinant of EL classifier's performance in CYP.