Wildfire Risk Prediction and Detection using Machine Learning in San Diego, California
Ashima Malik, Nasrajan Jalin, Shalu Rani, Priyanka Singhal, S. Jain, Jerry Gao
Abstract
Wildfires are the most deadly and dangerous accidents across the United States, especially in California. Many lives are lost, and billions of dollars’ worth of property damages occur in wildfire every year. Wildfires are fueled and accelerated by several different factors, such as weather, climate, vegetation types, land cover, and human activities. This research aims to develop a machine learning and big data-based fire risk prediction model considering the different geographical factors outlined above. We propose a systematic way to make fire risk prediction and detection models that analyzes satellite data, weather data, and historical fire data to predict fire. We got an accuracy of 100% using ensemble model and 93% using Faster R-CNN for wildfire risk prediction and fire detection, respectively, using machine learning models.