Every year, natural disasters kill around 90,000 people and affect 160 million people worldwide. There has been a general upward trend in natural disaster occurrences since 1980, most likely due to the worsening of climate change. With recent developments in machine learning and enormous amounts of natural disaster datasets, AI has immense capabilities in predicting the occurrence of numerous natural disasters, which can potentially save thousands of lives. There is currently no natural disaster predictor for the generalized public to easily access and use; thus, this project strives to use machine learning techniques to identify correlations in historical data of three major natural disasters to calculate the risk based on user input.
In this project, the three major natural disasters detected are earthquakes, forest fires, and hurricanes. By collecting user data of location and date when the user accesses the website, the Random Forest Regression model produces a depth and magnitude prediction of upcoming earthquakes. The user can also see information about previous natural disasters, such as their locations, dates, and severity, to see if they are in a historical region of natural disaster occurrences. The web interface uses Google Maps API to provide easy usage and improved functionalities.
Although this project was able to apply an easy to use application combined with machine learning models, there are many further steps for improvement. One future step is to use the earthquake prediction machine learning model and apply is similarly to forest fires and hurricanes by training on their respective datasets. A linear regression model was also constructed for drought predictions as well, which can be implemented with the web interface and Google API for the future. In addition, more natural disasters can be easily integrated with the existing framework and models, in order to account for a greater range of natural disaster predictions.