[2C5-OS-7b] OS-7 (2)

[2C5-OS-7b-04] Estimation of Human Attributes Using History of Migration Behavior in Commercial Facilities

〇Tsubasa Nakagawa¹, Natsuki Oka¹, Masahiro Araki¹, Kaito Oka² (1. Kyoto Institute of Technology, 2. Adinte Co., Ltd.)

Keywords:Clustering, Attribute estimation, Series of movement, Data preprocessing

The purpose of this paper is to classify customers in commercial facilities according to the data of their moving histories. We preprocessed the data to improve the clustering results and analyzed features of each cluster that accumulated on weekdays or holidays by using the K-means++ method (Arthur, 2007). As a result, it was found that when multiple clusters having the same characteristics appeared, they were separated by the stay time. And more, potential relationships between stores with different characteristics were shown. We intend to cluster a larger commercial facility's data and use the cluster information for ad delivery, so we are working on improving the clustering accuracy in the near future.