Multimodal Explainable Learning for Joint Crop-Yield and Price Forecasting
IOSR Journal of Computer Engineering · 2026-03 · 28 (3) · 63-77
DOI: 10.9790/0661-2803036377 ↗Kurz gefasst
Fachartikel in der Zeitschrift IOSR Journal of Computer Engineering, 2026-03. Verfasst von Vivek Kumar und 1 weiteren Autor:innen.
Abstract (Originalsprache)
The forecasting of crop yield and of market price is needed for procurement planning and price-risk management, yet the two are predicted separately, though shared weather and crop conditions drive both. Here the two are forecast together by a single model, a Modified CNN-LSTM with multi-head attention, demonstrated on wheat in Haryana, India. Four public sources are drawn on, namely the crop-yield statistics of DES; daily AGMARKNET prices; the NDVI, EVI, and NDWI indices from Sentinel-2 and MODIS via Google Earth Engine; and ERA5-Land weather data, aligned to a district-week grid. In early fusion, the features are encoded together; in late fusion, yield is predicted first, and its supply signals are passed to a separate price model. On the held-out 2025 season across 22 districts, yield yields an R² of 0.762 and a MAPE of 4.13% under late fusion, and price an R² of 0.821 and a MAPE of 4.34% under early fusion, outperforming XGBoost and Random Forest. Each forecast includes a bootstrapped 90% prediction interval and attributions from Integrated Gradients and LIME. Since every stage is parameterised by the crop calendar, satellite identifiers, and market codes, the pipeline extends to crops and regions where public data are available
Zitieren
APA
Kumar, V., & Chakre, R. R. (2026). Multimodal Explainable Learning for Joint Crop-Yield and Price Forecasting. IOSR Journal of Computer Engineering, 28(3), 63-77. https://doi.org/10.9790/0661-2803036377
Vancouver
Kumar V, Chakre RR. Multimodal Explainable Learning for Joint Crop-Yield and Price Forecasting. IOSR Journal of Computer Engineering. 2026;28(3):63-77. doi:10.9790/0661-2803036377
BibTeX
@article{Kumar2026,
author = {Kumar, Vivek and Chakre, Rahul R.},
title = {Multimodal Explainable Learning for Joint Crop-Yield and Price Forecasting},
journal = {IOSR Journal of Computer Engineering},
year = {2026},
volume = {28},
number = {3},
pages = {63--77},
doi = {10.9790/0661-2803036377}
}
EUVIMED