Stworzone rozwiązania zostały zaprezentowane w sesji poświęconej systemom wykorzystującym metody uczenia maszynowego i są wynikiem badań zrealizowanych przez Macieja Grzendę, Marcina Lucknera i Przemysława Wronę.
Zachęcamy do zapoznania się z abstraktem pracy „Urban Traveller Preference Miner: modelling transport choices with survey data streams” przedstawionej na konferencji:
The unprecedented interest in sustainable transport modes for urban areas raises the question of what makes citizens select environmentally friendly transport modes such as public transport rather than private cars. While travel surveys are conducted to document real transport mode choices, they can also shed light on how these choices are made.
In this paper, we demonstrate a system combining survey data with complex information documenting public transport features, as perceived by individual respondents. The system relies on a combination of big data modules to collect vehicle location records and travel planning engines to calculate candidate connection features, including disruptions faced by individuals. Hence a combination of streaming and batch modules is used to transform survey data into instances used to learn classification models. This takes place while taking into account concept drift. Real-life data from the city of Warsaw, including recently collected survey data, location records of trams and buses, and planned and true schedules, are used to demonstrate the system. A video related to this paper is available at https://youtu.be/fTcxUxEMGlk
