Files
Abstract
Weather prediction or forecasting is a science that has existed for a long time. There have been different methods recorded for these predictions with the use of tools like Numerical Weather Prediction. In the scientific community, there is a growing use of Machine Learning for this type of prediction and its ability to support large datasets also makes it a good alternative. We ran different machine learning models with the same dataset using different tools/frameworks which are Scalation, Scikit-Learn and Statsmodels. The result of this analysis is taken and compared against each other based on sample size, accuracy, machine learning models and the performance of the tools with respect to each other. The features or variables in the data are ranked to determine their level of importance when it comes to weather prediction. Some of the features present are temperature from weather stations, vegetation, and the density of man-made buildings. The accuracy is determined by the R2, MAE and RMSE values, respectively. With the machine learning models used the best accuracy was achieved with Random Forest and the tool with the fastest computation time was Scikit-Learn.