Meteorological data prediction over selected stations in Sub-Sahara Africa: Leveraging on Machine Learning Algorithm

This study investigated selected meteorological data prediction leveraging on a Machine Learning Algorithm Approach over five selected stations in Nigeria. The algorithm of Machine Learning was explored using weather parameters such as temperature, wind speed, wind direction and relative humidity to predict the rainfall rate. In the results, five Gaussian models (i.e., Rational Quadratic, Squared Exponential, Matern 5/2, Exponential and Optimized GPR) revealed different Root Mean Square Error (RMSE), Mean Square Error (MSE), and Mean Absolute Error (MAE) with prediction speeds ranging from 15000 to 26000 and the training time included 7.936, 1.8923, 2.3701, 3.267 and 282.19, respectively. The predicted response as against the true response for the two models shows a linear graph passing through the origin which confirmed a perfect regression model, where all the points lie on a diagonal line. Therefore, the relationship between MSE, MAE and RMSE for different models revealed that the optimized GPR has a better performance as compared to others. More so, visualizing the relationship between the output variable (rainfall) and each input variable reveals that some input variables (relative humidity, rainfall, pressure, wind speed and direction) have a strong correlation with the output variable (rainfall), with others having a noisy relationship which is not very clear.