Recommendation For Pick-up Points And Hot Zone Based On Taxi Trajectories

With the development of wireless communication technology and smart mobileterminal, recording mobile objects trajectories became quickly and easily. Thetrajectory information contains not only spatial attribute, but also temporal attribute,so it is major research objects in spatio-temporal mining technology, and itsapplication can provide location based service to usres, furthermore, it is also offercity planning, intelligent transportation etc.As taxi trajectories are easily to collecting, widely distributed and data usuallyvery big, these trajectories become the major research objects in spatio-temporalmining technology. Providing a recommended system service to taxis can helpsdrivers getting more income, alleviating traffic pressure caused by taxis cruising onthe road without a passenger, reducing unnecessary vehicle exhaust and promotingeffect on protecting environment. However, concentrated places with manypassengers always changed over time and city points of interests also have an impacton passengers’ travel trip. So, how to find out concentrated places in different time isa first and foremost condition in recommendation service.This paper first analyze trajectories dataset of mobile objects from MicrosoftResearch Asia and obtain the regular pattern of human activity with spatio-temporalchange, this regular pattern is different place have different attraction degree indifferent time interval. In order to get recommended place for taxi drivers indifferent time interval, it propose a method based on spatio-temporal clustering togenerate recommended places, this method is K-Means algorithm based onhierarchical clustering. It processed points of interests with density clustering toobtain human activity hotspots region and influence scope in different time interval.Finally, it combines the result of hotspots region and recommended places, providingthe recommended service to taxi drivers. It tested our method with the dataset whichprovided by Microsoft Research Asia and collected some service facilities (e.g.supermarket, parks) dataset by ourselves in Beijing. The result indicates that theserecommended places have a high accuracy rate, and represent the intensive degree ofpassengers in certain time interval.