Analysis of Taxi Drivers’ Behaviors
Analysis of Taxi Drivers’ Behaviors
In this paper, we study the historical trip and fare logs of Beijingtaxis to investigate the change of taxi drivers’ behaviors under moneypromotion.Beijing is the capital of the People’s Republic of China and isone of the most populous cities around the world. Its populationhad grown over 21 000 000 in 2013. However, there are only nearly67 000 licensed taxis in Beijing, which cannot meet the ever increasingdemand of taxis service, especially in rush hours.The dataset used in this paper includes about 8.3 million taxi tripsthat made by over 9000 taxis during 40 days. Each trip record includesthe pickup and drop-off location and time, as well as anonymized taxilicense numbers. The personally identifiable information of passengershas been properly anonymized, so that we can only differ the driversbut not the passengers. Moreover, the names of the drivers are notreleased, either.The longitude and latitude location information in each taxi trip isobtained by converting the received GPS data into a planar coordinatesystem, since the latitudinal trend is not pronounced in the central partof Beijing city. The location errors caused by inaccurate GPS signalsare much smaller than the moving distances of taxi trips and are thusomitted in this paper.As pointed out in many studies, human traveling activities can beinfluenced by many factors. For example, human traveling patterns arenotably different in working-days and weekends/ holidays. Moreover,the weather also greatly affects the calling amount of taxi services. Toreduce the influences of such factors, we choose the sampling daysas Oct. 16–18, 21, 22, 24, 25, 28–31, Nov. 18–22, 25–27, 2013 andFeb. 17, 18, 20, 21, 24–28, Mar. 17–21, 24–28, 2014, respectively. Wecall the sampling days in 2013 the first time period, in which no moneypromotion was applied. The sampling days in 2014 is called the secondtime period, in which money promotion was applied. All these days areworking days and the weathers in these days are mild so that the traveldemands in these days are similar siva sakthi travels
We examine four taxis service indices that correspond with thehypotheses mentioned in Section I:1) The distributions of the number of trips served by every vehicleper day during two periods, respectively;2) The distributions of the idle time lengths of every vehicle beforeper trip during two periods, respectively;3) The distributions of the traveling distances per trip during twoperiods, respectively;4) The spatial distributions of the origins and destinations of pas-sengers during two periods, respectively.Here, we choose the distributions of service data for comparison,since these distributions can provide an overall sketch of taxi serviceobserved in practices. Moreover, these distributions are estimatedfrom thousands of taxis and are thus robust to temporary behaviorvariations of some individual taxi drivers that may be caused byvarious disturbances (including abrupt vehicle failures).Fig. 1(a) shows the (sorted) numbers of trips made by every vehiclein each day of two periods. Each curve in Fig. 1(a) stands for the sortednumbers of trips made by every vehicle in a particular day. We cansee that, on average, drivers picked up significantly made more tripsunder the money promotion. The numbers of trips made by those mostdiligent drivers (whose indices are smaller than 2500) were roughlydoubled under the money promotion. gives the correspondinghistogram plot of the numbers of trips made by every vehicle per dayduring two periods.further compares the average numbers of passengers servedper half hour during two time periods. Clearly, taxis drivers servealmost 50% more passengers under the money promotion, during theworking hours (6:00A.M.–18:00P.M.). This result strongly supportsHypothesis 1 that drivers pick up more passengers under the moneypromotion.Fig. 3 shows the idle time lengths of every vehicle before per trip ineach day of two periods. We can see that there is a notable increase ofshort (less than 5 minutes) idle time lengths in the second time period;while the long-tail shape of the idle time length distributions remainunchanged. This is mainly because the long idle times are usuallycaused by unavoidable time breaks (e.g., break for lunch) of drivers.The money promotion cannot shorten the lengths of such time breaks.However, significantly larger occurrences of short idle times supportHypothesis 2 that the money promotion reduce the average time gapbetween two trips and thus the waiting time of passengers.Fig. 4 shows the distributions of the traveling distances per tripin each day of two periods. Here, the traveling distance of each tripis approximately calculated as the Euclidean distance between theoriginal and destination points. Since the layout of Beijing city is theresult of uniform planning and has a regular chessboard pattern, the es-timation error of traveling distance can be omitted.https://srisivasakthitravels.com/
https://www.facebook.com/Sri-Siva-sakthi-Travels-102008227873939/
In this paper, we study the historical trip and fare logs of Beijingtaxis to investigate the change of taxi drivers’ behaviors under moneypromotion.Beijing is the capital of the People’s Republic of China and isone of the most populous cities around the world. Its populationhad grown over 21 000 000 in 2013. However, there are only nearly67 000 licensed taxis in Beijing, which cannot meet the ever increasingdemand of taxis service, especially in rush hours.The dataset used in this paper includes about 8.3 million taxi tripsthat made by over 9000 taxis during 40 days. Each trip record includesthe pickup and drop-off location and time, as well as anonymized taxilicense numbers. The personally identifiable information of passengershas been properly anonymized, so that we can only differ the driversbut not the passengers. Moreover, the names of the drivers are notreleased, either.The longitude and latitude location information in each taxi trip isobtained by converting the received GPS data into a planar coordinatesystem, since the latitudinal trend is not pronounced in the central partof Beijing city. The location errors caused by inaccurate GPS signalsare much smaller than the moving distances of taxi trips and are thusomitted in this paper.As pointed out in many studies, human traveling activities can beinfluenced by many factors. For example, human traveling patterns arenotably different in working-days and weekends/ holidays. Moreover,the weather also greatly affects the calling amount of taxi services. Toreduce the influences of such factors, we choose the sampling daysas Oct. 16–18, 21, 22, 24, 25, 28–31, Nov. 18–22, 25–27, 2013 andFeb. 17, 18, 20, 21, 24–28, Mar. 17–21, 24–28, 2014, respectively. Wecall the sampling days in 2013 the first time period, in which no moneypromotion was applied. The sampling days in 2014 is called the secondtime period, in which money promotion was applied. All these days areworking days and the weathers in these days are mild so that the traveldemands in these days are similar siva sakthi travels
We examine four taxis service indices that correspond with thehypotheses mentioned in Section I:1) The distributions of the number of trips served by every vehicleper day during two periods, respectively;2) The distributions of the idle time lengths of every vehicle beforeper trip during two periods, respectively;3) The distributions of the traveling distances per trip during twoperiods, respectively;4) The spatial distributions of the origins and destinations of pas-sengers during two periods, respectively.Here, we choose the distributions of service data for comparison,since these distributions can provide an overall sketch of taxi serviceobserved in practices. Moreover, these distributions are estimatedfrom thousands of taxis and are thus robust to temporary behaviorvariations of some individual taxi drivers that may be caused byvarious disturbances (including abrupt vehicle failures).Fig. 1(a) shows the (sorted) numbers of trips made by every vehiclein each day of two periods. Each curve in Fig. 1(a) stands for the sortednumbers of trips made by every vehicle in a particular day. We cansee that, on average, drivers picked up significantly made more tripsunder the money promotion. The numbers of trips made by those mostdiligent drivers (whose indices are smaller than 2500) were roughlydoubled under the money promotion. gives the correspondinghistogram plot of the numbers of trips made by every vehicle per dayduring two periods.further compares the average numbers of passengers servedper half hour during two time periods. Clearly, taxis drivers servealmost 50% more passengers under the money promotion, during theworking hours (6:00A.M.–18:00P.M.). This result strongly supportsHypothesis 1 that drivers pick up more passengers under the moneypromotion.Fig. 3 shows the idle time lengths of every vehicle before per trip ineach day of two periods. We can see that there is a notable increase ofshort (less than 5 minutes) idle time lengths in the second time period;while the long-tail shape of the idle time length distributions remainunchanged. This is mainly because the long idle times are usuallycaused by unavoidable time breaks (e.g., break for lunch) of drivers.The money promotion cannot shorten the lengths of such time breaks.However, significantly larger occurrences of short idle times supportHypothesis 2 that the money promotion reduce the average time gapbetween two trips and thus the waiting time of passengers.Fig. 4 shows the distributions of the traveling distances per tripin each day of two periods. Here, the traveling distance of each tripis approximately calculated as the Euclidean distance between theoriginal and destination points. Since the layout of Beijing city is theresult of uniform planning and has a regular chessboard pattern, the es-timation error of traveling distance can be omitted.https://srisivasakthitravels.com/
https://www.facebook.com/Sri-Siva-sakthi-Travels-102008227873939/
Comments
Post a Comment