| dc.description.abstract | Agriculture plays a key role in Sri Lanka’s economy, especially in dry-zone areas
such as Anuradhapura, where soil conditions, water availability, and climate strongly
affect crop productivity. However, farmers often lack integrated, data-driven tools
to identify suitable cultivation areas and accurately predict yields. To address this
gap, this study develops and validates a GIS–AI-based framework for land suitability
assessment and yield prediction of chili (Capsicum annuum) in the Rajanganaya area.
The framework integrates GIS-based spatial analysis with machine learning techniques.
A weighted overlay analysis using soil, environmental, and climatic factors such as soil
type, pH, nutrients, slope, temperature, and water availability was used to generate
a land suitability map classified into five levels. For yield prediction, Random Forest
and Gradient Boosting models were trained using historical yield data, soil properties,
weather conditions, and fertilizer usage. The Random Forest model performed best,
achieving an R² value above 0.85 with low RMSE, indicating high prediction accuracy.
Results show that areas near irrigation tanks with balanced soil nutrients have very
high suitability and yield potential, while regions with poor soil fertility or limited
water access exhibit lower productivity. To support practical use, the validated model
was implemented in a mobile application that provides real-time yield predictions and
agronomic recommendations. Overall, the study demonstrates that integrating GIS and
AI can enhance precision agriculture, support sustainable farming, and enable data driven decision-making in Sri Lanka’s dry-zone agriculture, with potential applicability
to other crops and regions. | en_US |
