Il corso ha lo scopo di introdurre le principali tecniche di data mining e machine learning (incluso deep learning) per l'analisi di dati temporali, in particolare di time series e dati spazio-temporali relativi alla mobilita' umana. La presentazione delle nozioni sara' supportata da diversi casi di studio sviluppati dal laboratorio SoBigData.eu.
Time Series Analysis, Mobility data types; spatio-temporal trajectories; routines and mobility profiles; trajectory patterns and trajectory models.
The methods and techniques presented include the following: trajectory and profile reconstruction methods; O/D matrices of the mobility demand; trajectory clustering; flock-, swarm- and convoy- patterns; mobile activity recognition. In the course, the M-Atlas platform will be introduced, for the analysis of mobile objects' trajectories.
Sample case studies will be presented, on: study of urban mobility through GPS data; study of presence and inter-city fluxes through mobile phone data (application of Big Data for Official Statistics); traffic planning on national scale through mobile phone data; public transport studies (performances and integration with carpooling) based on GPS and planned public transport service. Exercises will involve (also) public GPS datasets (taxi) and/or social media data (Flickr).
Knowledge of the main opportunities, issues and limitations in the analysis of Bog Mobility Data; knowledge of the main data mining methods for such data; skills to use some specific analysis tools, developed over real datasets.
