Modeling of Spatial-Temporal Associations on a Mobile Trajectory

Despite the advanced state of development of mining association rules for time series data, simultaneous spatial and temporal data mining-particularly for mobile trajectories-remains a challenge, and requires huge amounts of data to be scanned during the mining process. This paper proposes a method that improves upon the incremental Godin algorithm, using a concept lattice to establish association rules based on spatial-temporal data with only a single scan of the dataset. Data was sourced from the mobile trajectories of individuals in a Tokyo subway station, then cleaned up, normalized, and compressed into one large locomotion dataset. Analysis of this data was executed using the newly proposed Indexed Lattice method, with results compared to those achieved by the traditional Godin and Apriori algorithms. It was shown that the Indexed Lattice method performed more efficiently and with greater stability, especially when processing data possessing more attributes.