Car pooling is expected to significantly help inreducing traffic congestion and pollution in cities by enablingdrivers to share their cars with travellers with similar itinerariesand time schedules. A number of car pooling matching serviceshave been designed in order to efficiently find successful ridematches in a given pool of drivers and potential passengers.However, it is now recognised that many non-monetary aspectsand social considerations, besides simple mobility needs, mayinfluence the individual willingness of sharing a ride, whichare difficult to predict. To address this problem, in this studywe propose GOTOGETHER, a recommender system for carpooling services that leverages on learning-to-rank techniquesto automatically derive the personalised ranking model of eachuser from the history of her choices (i.e., the type of acceptedor rejected shared rides). Then, GOTOGETHER builds the listof recommended rides in order to maximise the success rateof the offered matches. To test the performance of our schemewe use real data from Twitter and Foursquare sources in orderto generate a dataset of plausible mobility patterns and riderequests in a metropolitan area. The results show that theproposed solution quickly obtain an accurate prediction of thepersonalised user's choice model both in static and dynamicconditions.
A Machine-Learned Ranking Algorithm for Dynamic and Personalised Car Pooling Services
Campana MG;Delmastro F;Bruno R
2016
Abstract
Car pooling is expected to significantly help inreducing traffic congestion and pollution in cities by enablingdrivers to share their cars with travellers with similar itinerariesand time schedules. A number of car pooling matching serviceshave been designed in order to efficiently find successful ridematches in a given pool of drivers and potential passengers.However, it is now recognised that many non-monetary aspectsand social considerations, besides simple mobility needs, mayinfluence the individual willingness of sharing a ride, whichare difficult to predict. To address this problem, in this studywe propose GOTOGETHER, a recommender system for carpooling services that leverages on learning-to-rank techniquesto automatically derive the personalised ranking model of eachuser from the history of her choices (i.e., the type of acceptedor rejected shared rides). Then, GOTOGETHER builds the listof recommended rides in order to maximise the success rateof the offered matches. To test the performance of our schemewe use real data from Twitter and Foursquare sources in orderto generate a dataset of plausible mobility patterns and riderequests in a metropolitan area. The results show that theproposed solution quickly obtain an accurate prediction of thepersonalised user's choice model both in static and dynamicconditions.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
