The Prediction of the Impact of Climate Change on Agricultural Productivity in Kenya Using Machine Learning

Kimani, Samuel Mbugua (2026) The Prediction of the Impact of Climate Change on Agricultural Productivity in Kenya Using Machine Learning. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6026241027-Master_Thesis.pdf Restricted to Repository staff only Download (1MB)

Abstract

Climate change is a significant challenge to agricultural productivity in Kenya because crop production is strongly influenced by rainfall, temperature, vegetation conditions, and other environmental pressures. Although machine learning is increasingly used for agricultural yield prediction, many studies still emphasize predictive accuracy and provide limited explanation of the climate variables influencing predictions. This study develops a Climate Decision Support System, an explainable machine-learning framework for predicting the impact of climate change on agricultural productivity in Kenya. Rainfall and temperature data are obtained from the World Bank, while crop yield, pesticide use, and Normalized Difference Vegetation Index (NDVI) indicators are obtained from Food and Agriculture Organization of the United Nations (FAO). The study focuses on five representative crops: sugar cane, potatoes, tomatoes, bananas, and avocados. The methodology includes data integration, climate trend analysis, representative crop selection, feature engineering, statistical testing, machine learning modelling, explainable AI, climate scenario simulation, recursive forecasting, vulnerability ranking, and interactive dashboard development. The models include Linear Regression, Bayesian Ridge Regression, Random Forest, Gradient Boosting, XGBoost, and CatBoost. Climate-only models are used for climate-impact interpretation, while climate-plus-lag models are used for operational forecasting. The results indicate a statistically significant warming trend during the modelling period, while long-term rainfall patterns are weaker and less consistent. Model performance varied across crops, with the climate-plus-lag Gradient Boosting model achieving the lowest RMSE for potatoes (RMSE = 24,651.295 hg/ha). Climate-productivity relationships differ across crops, confirming the importance of crop-specific modelling.
================================================================================================================================
Perubahan iklim merupakan tantangan besar bagi produktivitas pertanian di Kenya karena produksi tanaman sangat dipengaruhi oleh curah hujan, suhu, kondisi vegetasi, dan tekanan lingkungan lainnya. Meskipun pembelajaran mesin semakin banyak digunakan untuk memprediksi hasil pertanian, banyak penelitian masih berfokus pada akurasi prediksi dan hanya sedikit yang menjelaskan variabel iklim yang memengaruhi hasil prediksi. Penelitian ini mengembangkan Climate Decision Support System, yaitu kerangka pembelajaran mesin yang dapat dijelaskan untuk memprediksi dampak perubahan iklim terhadap produktivitas pertanian di Kenya. Data curah hujan dan suhu diperoleh dari Bank Dunia, sedangkan data hasil tanaman, penggunaan pestisida, dan indikator Normalized Difference Vegetation Index (NDVI) diperoleh dari Organisasi Pangan dan Pertanian Perserikatan Bangsa-Bangsa (FAO). Penelitian berfokus pada lima tanaman representatif, yaitu tebu, kentang, tomat, pisang, dan alpukat. Metodologi penelitian mencakup integrasi data, analisis tren iklim, pemilihan tanaman representatif, rekayasa fitur, uji statistik, pemodelan pembelajaran mesin, penerapan AI yang dapat dijelaskan, simulasi skenario iklim, peramalan rekursif, pemeringkatan kerentanan, dan pengembangan dasbor interaktif. Model yang digunakan meliputi Regresi Linear, Regresi Bayesian Ridge, Random Forest, Gradient Boosting, XGBoost, dan CatBoost. Model khusus iklim digunakan untuk menginterpretasikan dampak iklim, sedangkan model iklim-plus-lag digunakan untuk peramalan operasional. Hasil penelitian menunjukkan adanya tren pemanasan yang signifikan secara statistik selama periode pemodelan, sedangkan pola curah hujan jangka panjang menunjukkan hubungan yang lebih lemah dan kurang konsisten. Kinerja model bervariasi antar tanaman; model Gradient Boosting iklim-plus-lag mencapai RMSE terendah untuk kentang, yaitu sebesar 24.651,295 hg/ha. Hubungan antara iklim dan produktivitas berbeda pada setiap tanaman, sehingga menegaskan pentingnya penerapan pemodelan yang spesifik terhadap karakteristik masing-masing tanaman.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	agricultural productivity, climate change, decision support system, explainable AI, machine learning
Subjects:	T Technology > T Technology (General) > T58.62 Decision support systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 59101-(S2) Master Thesis
Depositing User:	Samuel Mbugua Kimani
Date Deposited:	11 Aug 2026 09:13
Last Modified:	11 Aug 2026 09:13
URI:	http://repository.its.ac.id/id/eprint/144310

Available Versions of this Item

• The Prediction of the Impact of Climate Change on Agricultural Productivity in Kenya Using Machine Learning. (deposited 10 Aug 2026 01:05)
  • The Prediction of the Impact of Climate Change on Agricultural Productivity in Kenya Using Machine Learning. (deposited 11 Aug 2026 09:13) [Currently Displayed]

Actions (login required)

View Item