Hurricane Damage Detection using Convolutional Neural Network and Customized KNN

Hurricane hits have great harm to people's lives, and it is very important to provide help to the affected area in time after the hit. This paper proposes a model to predict whether a hurricane damages a house through satellite images. I apply a logistic regression model and two convolutional neural network models and find the AlexNet's best performance. To use the location information of the images, I make certain modifications to the KNN model and combine it with AlexNet for hurricane damage detection and classification. I find that the new model has the best classification result, with an accuracy rate of 95.39% and an F1 value of 0.9739. The model-based method can better help relevant government departments and provide timely and accurate assistance to the disaster-stricken areas after the hurricane hits.