Predictive Analysis of Crop Yield Based on Environmental and Soil Conditions

Abstract

Accurate crop yield prediction is critical for sustainable agriculture, enabling better planning and resource allocation. Crop yield depends on various factors, including environmental and soil conditions, making it a complex prediction task. This paper presents a predictive analysis framework that leverages machine learning models to estimate crop yield based on environmental factors (temperature, rainfall, humidity) and soil properties (pH, nutrients, moisture). This study compares two distinct predictive models—linear regression and decision trees—to determine which method provides better accuracy and interpretability for yield prediction. The primary aim is to explore the correlation between these key factors and crop yield outcomes, based on real-world data from an experimental farm. The study applies feature engineering techniques to preprocess environmental and soil da-tasets, followed by model training and validation using cross-validation techniques. The results of this analysis provide insights into the most critical factors influencing crop yield and offer a comparative performance analysis between the two machine learning models. This research demonstrates that decision trees outperform linear regression in terms of accuracy but highlight areas where linear regression could still be valuable for interpretability.