Predicting the Next Country Considering Features of Tourism Types

Abstract

We propose a method for predicting traveler's next countries using the LSTM deep learning method. To predict the next country accurately, it needs to consider features of tourism that involve travel experiences, in addition to country information such as location information and culture of each country. Therefore, we propose a method considering country information using Wikipedia and features of tourism types, obtained through analyzing social media with Random Forest. In this paper, we conducted an experiment using LSTM with Wikipedia and tourism types of input data, and obtained the results, which were better than baseline methods, of Accuracy@1, Accuracy@3, Accuracy@5, Accuracy@10, and MRR scores of 26.06%, 46.31%, 56.61%, 70.30%, and 0.403, respectively. Furthermore, the effectiveness of considering tourism types was confirmed at the significance level of 5%.