Towards Efficient Discovery of Spatially Interesting Patterns in Geo-referenced Sequential Databases

A geo-referenced time series is a crucial form of spatiotemporal data. Useful information that can empower the users to achieve economic development is hidden in this series. When confronted with this problem, researchers modeled this series as a transactional database and discovered various user interest-based patterns. Since transactional databases disregard the items' sequential ordering information, existing studies are inadequate to find interesting patterns in the data of those applications, where the items' sequential ordering needs to be considered. With this motivation, this paper first presents a new data transformation technique that converts geo-referenced time series data into a geo-referenced sequential database that preserves the items' sequential occurrence information. Second, this paper presents a novel model of geo-referenced frequent sequential patterns that may exist in a database. Third, a novel neighborhood-aware exploration technique has been presented to effectively reduce the search space and the computational cost of finding the desired patterns. Finally, we present an efficient algorithm to find all desired patterns in a database. Experimental results demonstrate that the proposed algorithm is efficient. We demonstrate the usefulness of our patterns with a case study, which involves finding congestion patterns in road network data.