FISH PRODUCTION MODELING AND FORECASTING IN INDIA USING THE XGBOOST ALGORITHM Authors: Shikha Yadav, Binita Kumari, Divya Sharma, Abdullah Mohammad Ghazi Al khatib, Bayan Mohamad Alshaib, Yashpal Singh Raghav, Promil Kapoor, Tufleuddin Biswas, Soumik Ray, Aditya Pratap Singh, Pradeep Mishra Journal: Journal of Animal and Plant Sciences (JAPS) ISSN: 1018-7081 (Print), 2309-8694 (Online) Volume: 35 Issue: 2 Pages: 522-530 Year: 2025 DOI: https://doi.org/10.36899/JAPS.2025.2.0043 URL: https://doi.org/https://doi.org/10.36899/JAPS.2025.2.0043 Publisher: Pakistan Agricultural Scientists Forum Abstract:

Time series analysis using machine learning is vital for forecasting in commodity sciences. This research leverages advanced machine learning models for time series forecasting of fish production at both state and national levels in India. The study developed and compared traditional models, like the autoregressive integrated moving average (ARIMA) and state space models, with the advanced machine learning model, extreme gradient boosting (XGBoost), using training and test data sets. Results showed that XGBoost and state space models significantly outperformed the ARIMA model. Specifically, XGBoost had the highest accuracy in eight of eighteen series, followed by state space models (seven out of eighteen), and ARIMA models (three out of eighteen). This confirms that applying diverse machine learning models can enhance forecasting accuracy for fish production. After identifying the best-performing models, forecasts for fish production were extended to 2030, indicating that India’s total and marine fish production would likely continue to grow, with minimal change expected in key producing states. This data-driven analysis offers valuable insights for food security planning and policy-making in the region.

Keywords: Fish production, Time series, machine learning, forecasting, ARIMA, XGBoost