Next place prediction in unfamiliar places considering contextual factors
Takashi Nicholas Maeda(the University of Tokyo), Kota Tsubouchi and Fujio Toriumi (the University of Tokyo)
ACM SIGSPATIAL 2017, 2017/11
データサイエンス (Data Science)
- This research aims to develop a method that maximizes the accuracy of next place prediction (NPP) in places that are unfamiliar to each mobile phone users. If it becomes possible to predict the next place in unfamiliar places, it will be possible to present the detour to the user of the smartphone in advance, when there is an accident, a construction, or congestion on the way to the next place. In such places, it is difficult to predict the next place based on each person’s historical location data because there are just a few or no data in such places for each user. Furthermore, there is also difficult to rely on the regularity of human mobility, because tourists’ mobility is easily affected by other factors such as weather. Our research aims to solve the difficulties in NPP in unfamiliar places by focusing on contextual factors such as weather, transportation means, place of residence, and time. We divide mobility data of mobile phones according to the values of such contextual factors. Then, our method learns and predicts next places for each piece of data divided. The most suitable combination of factors to divide data is automatically chosen. In our evaluation, it is shown that our method can successfully predict next places in unfamiliar places. In this paper, we also show that location data of mobile phone with few missing records are crucial for our method.