Development of the Similar Typhoon Search System Based on the Deep Neural Network Using Deep Learning

We present a method for improving the search accuracy of the similar typhoon research system by leveraging deep neural networks, which are an extension of artificial neural networks. To the search engine, we apply three parameters of typhoons: course, temporal central pressure, and speed. We show that these parameters can improve accuracy when searching for past typhoons having characteristics similar to the target. Furthermore, the accuracy of functions designed to support the expected disaster prevention actions and flood fighting services was assessed based on the results from the search system.