Ευρετηρίαση δεδομένων κίνησης με χρήση μοντέλων
Model - based indexing of mobility data
Ηλιάκης, Αρτέμης Ν.
During the last decade, the domain of mobility data management and mining has emerged, providing many effective methods for mining intuitive patterns that represent collective behavior of trajectories of moving objects. An interesting research line is that instead of operating on the raw data collected from various sensors, researchers make use of semantically enriched data, which are either declared (by the users) or inferred (by some annotation method). This way raw trajectory data is transformed into spatio-temporal-textual sequences, where the extra textual information that is added to the dimensions of space and time, represent the movement semantics. Such spatio-temporal-textual sequences form a more realistic representation model of the complex every-day life (and as such the mobility) of individuals. The last years, there is a mainstream using stochastic models in information retrieval systems of sequential data. Specifically, a particular type of Markov models, named Hidden Markov Models (HMMs) have been successfully applied in speech recognition, music pattern recognition, consumer pattern recognition and many other domains of sequential data. Aiming to achieve high accuracy in indexing mobility data we will apply a model-based classification by representing each class of a mobility database via a HMM.