- Version
- Download 1
- File Size 308.19 KB
- File Count 1
- Create Date September 3, 2026
- Last Updated September 3, 2026
LOCALIZED MACHINE LEARNING-BASED WEATHER PREDICTION MODEL FOR SMALLHOLDER FARMERS IN RURAL NIGERIA
ABSTRACT
This paper presents the design of a localised machine learning-based framework for weather forecast model which aims at assist farmers in rural areas towards making climate-sensitive decisions in order to work effectively. The historical climate data that was used for the system implementation were gathered and analysed using the data cleaning, normalisation, time-series structuring, and feature engineering processes (lag variables and seasonal indicators) which was based on temperature and rainfall data found in the World Bank Climate Change Knowledge Portal. To provide consistency in time and compute efficiency, a time-conscious data splitting method was used to implement a Linear Regression algorithm and to train the algorithm in the Google Colab environment. The model’s performance was evaluated using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and the coefficient of determination (R2 score) and the implementation results showed strong predictive capability for temperature forecasting, achieving an MAE of 0.92oC, RMSE of 1.25oC, and R2 score of 0.84, indicating high explanatory power and reliable trend capture.
Keywords: Localized Weather Prediction, Machine Learning, Linear Regression, Climate Data, Temperature Forecasting