Indian Green Coffee for Enhancing Export Predictions Using Advanced Algorithms for and Contextual Data

Abstract

This study aims to improve the accuracy of forecasting Indian green coffee exports by using advanced predictive models and incorporating external factors like climate conditions and geopolitical influences. It addresses gaps in previous research by combining ARIMA and Random Forest models, enhanced with global coffee demand and climate data. The research uses a detailed dataset of Indian green coffee exports, stored in MongoDB, with preprocessing steps to ensure data consistency. The methodology integrates ARIMA (and SARIMA) with machine learning models, enriched by external variables. Among the models evaluated, Neural Networks achieved the best accuracy (MAE: 35.67, RMSE: 61.34), followed by Gradient Boosting (MAE: 37.89, RMSE: 62.45). Random Forest performed well, while ARIMA showed moderate results. The study highlights the importance of advanced models and external data for more precise coffee export predictions, offering valuable insights for Indian exporters and contributing to agricultural predictive analytics.