Unsupervised Top ographic Learning for Spatiotemp oral Data Mining

In re cent years, the s ize and complexit y of datasets have show n an exponential g row th. In m any application areas, huge amounts ofdata are gener ated, explicitly or i mplicitly containing spatial or s patiotempor a l infor mation. Howe ver, t he abilit y to a nalyze thesedata remains inadequate, and t he need for a dapted data mining to ols becomes a major challenge. In t his paper, we prop ose a newunsuper v i sed a lgor ithm, suitable for t he analysis of noisy s patiotempor al R adio Frequency I Dentification ( RFID) d ata. Two realapplications show that this algor i thm i s an e fficient d ata-mining to ol for b ehav ior al s tudies b ased on RFID te chnolog y. It allowsdiscover ing and compar ing stable patter ns in an RFID sig n al and i s suitable for continu ous l ear ning .