Takahiko KUSAKABE Assistant Professor, Division of Joint Usage and Research (2016-)
Further information available at http://www.csis.u-tokyo.ac.jp/~t.kusakabe/index_en.html

Recent advances of information technologies (such as IoT: Internet of Things) enable us to collect massive dynamic spatial information relating to human and automobile behavior. His study topics are relating to develop methodologies for implementing such advanced technologies for urban transport network and travel behavior analysis by using data-fusion, datamining, and machine-learning methods.

Understanding human behavior from massive dynamic spatial information

This study introduced visualization method and datamining analysis of large historical dataset, which is related to transportation systems such as traffic detector data collected in urban expressway, transit smart card data of public transport, and probe vehicle data. By using the proposed methods, this study revealed spatiotemporal characteristics of transport systems. Furthermore, model of travel patterns and their variability over long-term periods were developed to understand changes in travel demand. In the future study, simulation models based on such massive dataset will be investigated in order to estimate and predict human behavior in urban transport network.

Developing data fusion method

In the field of transport engineering, recently, passive dataset, such as traffic detector data, smart card data, probe vehicle data, Wi-Fi and Bluetooth, can be automatically and continuously collected along with operation of the systems. Most of these data provide continuous and long-term travel information which is difficult to achieve with a survey. However, they are fragmentary for analyzing human behavior. On the other hand, conventional dataset of human travel behavior is usually collected by designed surveys. They collect sufficient information about travels for analyzing travel behavior. But it is difficult to conduct continuous and long-term surveys due to their cost and burden of respondents. To overcome such shortcomings of each dataset and to fuse advantages, this study develops data fusion method using transit smart card and household survey data. The empirical data mining analysis showed that the proposed methodology can be applied to find and interpret the behavioral features observed in the smart card data which had been difficult to obtain from each independent dataset. In the future study, methodologies to fuse larger dataset (e.g. probe vehicle data) with various spatial information (e.g. weather data) will be investigated.