Research

For the purpose of accelerating the use of data in the era of Society 5.0, we will make efforts to develop a cross-data analytis technology for utilizing data obtained from a variety of sensing technologies and different kinds of social big data to construct a platform that will help develop and expand smart services with a view towards a sustainable society, and will conduct activities for accelerating open innovation using regional IoT platforms.

We are developing data mining technology for collecting and integrating sensing data accumulated at multiple data sources on demand for analyzing their association. It includes an integrated method of spatio-temporal clustering and association rule discovery for finding locally-associative datasets, an algorithm for faster processing of association rule mining in rare cases such as accidents and disasters, and an efficient method for high-utility frequent itemset mining. These technologies are applied, for example, to predict a risk of traffic disruptions such as accidents and congestion in the event of heavy rain, heavy snow or other abnormal weather conditions based on past cases.

We are developing prediction technology for meteorological, atmospheric and other environmental data based on the Convolutional Recurrent Neural Network (CRNN), which integrates spatial pattern learning by the Convolutional Neural Network (CNN) with temporal pattern learning by the Long Short Term Memory (LSTM). This technology is applied, for example, to short-term prediction of air quality index affected by trans-border air pollution. In this technology, we have sought an optimal method of integration between the CNN and the LSTM in consideration of the spatio-temporal continuity of environmental data to achieve high levels of versatility and prediction accuracy. We also endeavor to develop technology for optimizing prediction in collaboration between a variety of observation data and application-specific data collected from users.