Enhancing Predictive Accuracy for Agricultural Crop Yields in Indian States Using Power Transformation in Machine Learning Models

Abstract

Agricultural productivity is critical for the sustenance and economic growth of India, a country with diverse climatic regions and crop practices. Accurate prediction of crop yields is essential for optimizing agricultural outputs, ensuring food security, and planning resource allocation. Traditional predictive models often struggle with the nonlinear and complex nature of factors influencing crop yields. This study introduces an innovative approach by integrating power transformations, specifically the Yeo-Johnson transformation, within machine learning models to enhance the predictive accuracy of agricultural crop yields in various Indian states. We collected a comprehensive dataset comprising climatic variables, soil properties, crop types, and historical yield data across different Indian states. The Yeo-Johnson power transformation was applied to normalize the distribution of these features, addressing issues of skewness and heteroscedasticity, thereby making the data more suitable for machine learning algorithms. We evaluated several machine learning models, including Linear Regression, Random Forest, and Gradient Boosting, to establish a benchmark for comparison. Our findings demonstrate a significant improvement in predictive accuracy with the application of the Yeo-Johnson transformation across all models tested are shown in table 1. The Gradient Boosting model, post-transformation, exhibited the highest accuracy, underscoring the potential of combining power transformations with ensemble learning techniques for crop yield prediction. This approach not only aids in achieving greater precision in forecasts but also provides insights into the relative importance of various factors affecting crop yields. The study emphasizes the role of advanced data preprocessing techniques, such as power transformations, in unlocking the full potential of machine learning for agricultural applications. By enhancing the predictive accuracy of crop yield models, stakeholders can make more informed decisions, leading to improved agricultural productivity and sustainability in the diverse and challenging landscape of Indian agriculture.

Topics & Concepts