Applying Machine Learning Methods for Predicting Sand Storms

摘要

Natural disasters arise in different places in the world and vary depending on the structure and weather conditions of the area. A sandstorm is a strong wind that blows sand and dirt from a dry surface. Saudi Arabia has faced many sandstorms over the years, especially in the east and central region of the kingdom. It is apparent that the number of sandstorms is increasing every year. Sandstorms can cause serious problems such as lack of visibility and breathing problems. The aim of this research is to predict if a sandstorm is going to appear up to 24 hours ahead in real-time by using an appropriate machine learning methods and displaying the results to the user through a simple interface. Our model serves users in three cities in Saudi Arabia. In this research, we investigate which models perform better in predicting sandstorms. The best performing model is used in our website. Our results show that CART decision tree outperforms naive Bayes and logistic regression.