Rainfall Prediction Models for Katsina State, Nigeria: Machine Learning Approach

Abstract

Weather patterns and rainfall are essential pieces of information that drive the agricultural sector. For a peasant farmer in katsina, knowledge of pattern of rainfall is a vital determinant of which crops to plant and when to commence planting. Considering its implications for agriculture, water resource management, and disaster preparedness, this paper developed rainfall prediction models specifically tailored to the unique patterns in Katsina State, Nigeria. The work used Multiple Linear Regression (MLR) and Artificial Neural Network (ANN) with 10 years historical rainfall data obtained from the Nigerian Meteorological Agency Katsina where data were collected from Nigerian Meteorological Agency Katsina State, then preprocessed and subjected to the models, while the evaluation matrices that were used are Precision, Recall, Accuracy, F1-Score, R- Square, Mean Square Error (MSE) and Root Mean Square Error (RMSE). The results of this research indicate that the ANN model outperforms the MLR model with R-Square of ANN equal 0.532 and that of MLR equal to 0.099 and also the precision, recall and f1-score for ANN are 0.666, 0.666 and 0.661 respectively while for MLR they are 0.580, 0.583 and 0.564 respectively. These findings suggest that the ANN model is better at capturing the linear relationship between input variables and rainfall.