Logo PTI Logo FedCSIS

Proceedings of the 21st Conference on Computer Science and Intelligence Systems (FedCSIS)

Annals of Computer Science and Information Systems, Volume 47

Ridge Regression with Walk-Forward Validation and Explainable AI for Tea Yield Prediction in Sri Lanka

DOI: http://dx.doi.org/10.15439/2026F5989

Citation: Kavishka Karunagaran (). Ridge Regression with Walk-Forward Validation and Explainable AI for Tea Yield Prediction in Sri Lanka. In M. Bolanowski, M. Ganzha, M. Grzegorowski, L. Maciaszek, M. Paprzycki, A. Paszkiewicz, D. Ślęzak (eds), Proceedings of the 21st Conference on Computer Science and Intelligence Systems (FedCSIS). ACSIS, Vol. 47, pages 543–548.

Full text

Abstract. Sri Lanka's tea industry has faced significant yield declines due to climate variability and a lack of data-driven forecasting tools. This paper presents a machine learning-based yield prediction system using Ridge Regression with lag feature engineering and walk-forward validation on longitudinal data from a Sri Lankan regional estate (2017--2025). The system integrates fertilizer records with historical weather data from the Open-Meteo Archive API to forecast monthly Yield Per Hectare (YPH). Comparative experiments empirically established an optimal 3-year rolling training window. Temporally ordered walk-forward validation is utilized to eliminate data leakage and realistically simulate agricultural forecasting conditions. Model explainability is achieved intrinsically via direct Ridge coefficient analysis. The proposed model achieves an R^2 of 0.9201, outperforming CatBoost, XGBoost, Random Forest, and Neural Networks. While validated on a single estate, results confirm its robustness for deployment within Sri Lankan tea estate management systems.

References

  1. 1] Asian Dispatch, “Sri Lanka tea production down 15% in April over scorching sun,” Asian Dispatch, May 26, 2024. [Online]. Available: https://asiandispatch.net/ global-tea-demand-booms-south-asian-workers-fight-climate-change/. [Accessed: Jun. 19, 2026].
  2. Central Bank of Sri Lanka, “Agriculture Sector Data Bulletin – December 2024,” CBSL, Colombo, Sri Lanka, 2024. [Online]. Available: https://www.cbsl.gov.lk/sites/default/files/cbslweb_documents/statistics/ agriculture_data_bulletin_2024_december_e.pdf. [Accessed: Jun. 19, 2026].
  3. P. Weththasinghe and U. I. Dissanayeke, “Alternative and ICT-based advisory and extension approach in the tea sector in Sri Lanka: An explorative study,” Tropical Agricultural Research, vol. 35, no. 1, pp. 46– 60, 2024. http://dx.doi.org/10.4038/tar.v35i1.8703
  4. L. Mampitiya, H. S. Sumanasekara, N. Rathnayake, Y. Hoshino, and U. Rathnayake, “Explainable artificial intelligence to estimate the Sri Lankan (Ceylon) tea crop yield,” Smart Agricultural Technology, vol. 11, p. 100999, 2025. http://dx.doi.org/10.1016/j.atech.2025.100999
  5. P. Deka, N. Tanti, and P. Neog, “Assessing predictive models for tea yield: A statistical and machine learning approach in Assam’s Biswanath Chariali District,” Journal of Experimental Agriculture International, vol. 46, no. 7, pp. 526–534, 2024. http://dx.doi.org/10.9734/jeai/2024/ v46i72605
  6. D. Batool et al., “A hybrid approach to tea crop yield prediction using simulation models and machine learning,” Plants, vol. 11, no. 15, p. 1925, 2022. http://dx.doi.org/10.3390/plants11151925
  7. P. Phan, N. Chen, L. Xu, and Z. Chen, “Using multi-temporal MODIS NDVI data to monitor tea status and forecast yield: A case study at Tanuyen, Laichau, Vietnam,” Remote Sensing, vol. 12, no. 11, pp. 1814, 2020. http://dx.doi.org/10.3390/rs12111814
  8. S. J. J. Jui et al., “Spatiotemporal hybrid random forest model for tea yield prediction using satellite-derived variables,” Remote Sensing, vol. 14, no. 3, p. 805, 2022. http://dx.doi.org/10.3390/rs14030805
  9. Q. Huang et al., “Automatic construction of knowledge graph of tea diseases and pests,” Annals of Computer Science and Information Systems, vol. 37, pp. 141–146, 2023. http://dx.doi.org/10.15439/2023F6100
  10. M. Aldawoud and Z. Sbaï , “Multi-level optimization-based ensemble machine learning for efficient crop yield prediction in Saudi Arabia,” Annals of Computer Science and Information Systems, vol. 43, pp. 429– 440, 2025. http://dx.doi.org/10.15439/2025F5652
  11. Open-Meteo, “Open-Meteo historical weather archive API,” 2024. [Online]. Available: https://open-meteo.com/en/docs/historical-weather-api. [Accessed: Jun. 19, 2026].
  12. A. E. Hoerl and R. W. Kennard, “Ridge regression: Biased estimation for nonorthogonal problems,” Technometrics, vol. 12, no. 1, pp. 55–67, 1970. http://dx.doi.org/10.2307/1267351
  13. F. Pedregosa et al., “Scikit-learn: Machine learning in Python,” Journal of Machine Learning Research, vol. 12, pp. 2825–2830, 2011. [Online]. Available: https://www.jmlr.org/papers/volume12/pedregosa11a/ pedregosa11a.pdf. [Accessed: Jun. 19, 2026].