The Hot Spots Of The Taxi Based On Data Mining And The Analysis Of The Traffic Flow Of Intersections In Nanjing

Since the 20th century,cities have developed rapidly and the urban space is also expanding at the same time.The development of city traffic is gradually becoming a judge of the development scale and potential of cities.Nowadays,the urban traffic congestion problem often focused on hot spots.How to identify the hot spots of residents' travel and dig behavior characteristics from the space and time are gradually taken seriously.However,former studies often use statistical research in the form of questionnaire,so there are many limitations and one-sidedness.AS the wireless devices with GPS and 3s integration technology developed rapidly,various kinds of sports mode of query and the collection and tracking of moving objects data have became more and more convenient.Simultaneously,the acquisition and storage of GPS data is more and more quick and convenient so that transportation has entered a peak.At present most of the cabs in the city has installed the GPS wireless devices and recorded the rich information of residents.Using the method of spatio-temporal data mining for GPS data track points to explore the implicit travel behavior pattern and solve the problem of road congestion transportation planning can can provide a scientific reference for the related departments which is about location-based services.According to the working days and weekend,the paper analyses and researchs the taxi GPS data to carry out effectively.The author divides different times from the day so as to study the regular characteristic of taxis' operation,at the same time discovers the hot spots of residents on the space leading to establishing traffic flow mode and analyses it in these areas,which have intersection.Above all can guide the taxi driver get the intelligent navigation,at the same time residents can avoid the traffic jams during the long wait.Finally it can also provide some forcing support for intelligent transportation.Therefore,this article mainly has carried out the following work:(1)For the pretreatment of taxi track points,from the angle of load factors and empty loading ratio,the author analyzed the total number of travel,each hour travel times9 passenger/idle time and mileage so as to get residents' temporal travel rule:On working day,it's obviously to find morning and evening peak;However there is not peak evidently on weekend,and the number of travel remains at a higher level.Whether working day or weekend,Passenger time is mainly within 20 minutes,the idle time is concentrated in the 10 minutes.This evidence proof that it is convenient to get on a taxi and the company has a good operating level too.But the mileage on working days is generally lower than holidays,it can also reflect the residents' travel which is no longer a single commuter travel,and they have added more travel for the purpose of entertainment.(2)A method based on exploratory spatial data analysis is proposed for urban residents to fi nd out travel hot spots.In this paper,with the help of the currently popular spatial autocorrelation technology,the author applied it to the urban residents' travel behavior research for discovering the macroscopic statistical law,and using the 3d image to contrast.At the same time the author used k-means method to find out the cluster centers of travel and contrast the interseccion in hot areas.Finally it is found out the relation,which is combined the intersection in hot spots and the cluster centers.The major areas is focused on especially in Jing Yi road,Jing Er road,Jing San road,Wei Si road and Wei Wu road.(3)Using the multiple time scale analysis method and the weighted average method which using the degree of variation factor as the weight,this article analysis the intersection traffic flow characteristics from the perspective of macro and micro based on the hot spots,the results show that volatility of working day is higher than day off.Studying the characteristics of traffic flow is of great significance to understand the traffic flow.Through the comparison of PSO-SVM model,the traditional BP neural network and SVM model,the correlation coefficient R is largest and prediction error is lower than others?PSO-SVM model is more suitable for traffic flow prediction...