Tourism is one of the main sources of GDP (Gross Domestic Product) in Taiwan. Web searching for internet data using computers and mobile phones is becoming more and more common. It has become indispensable for people to arrange tourism activities. We believe Google search engine data records could reflect internet user's preference for travel trends. In past researches, traditional time series SARIMA models and machine learning method have been used to build models. This study collects each month's numbers of Japanese and Korean tourists to Taiwan from 2011 to 2018, plus the corresponding Google keyword trend data in the same period. We compare the prediction effectiveness of the SARIMA model and the long short-term memory models of deep learning, to explore the effectiveness of deep learning in the application of time series data.