A Ride Route Recommendation Method Based On Traffic Congestion And In-car Crowdness Prediction

With the rapid development and popularization of mobile communication technology,more and more mobile intelligent terminals are applied to urban traffic management and planning,resulting in massive and extremely rich traffic big data.The traffic big data includes GPS trajectory data and IC card swiping data,GPS trajectory data is a serious of the real bus driving path points,which are collected with a certain sampling frequency,the IC card swiping data records the time and location information of the passengers when they swipe their cards.These data contain abundant spatio-temporal information of mobile objects,reflect the operation rules of urban public transport system and the travel characteristics of urban residents,which is of great significance for the research of adjustment and optimization of urban public transport lines and passenger travel route choice.Currently,the choice of passenger travel route is a difficulty in urban service research.More and more urban residents are pursuing better traffic service.How to mine the useful information from a large number of GPS trajectory data and IC card swiping data is conducive to improving urban traffic services and providing scientific basis for the choice of passenger travel route,which has become the key demand of traffic big data research.So far,the passenger travel route recommendation algorithms are mainly divided into three categories: the first is to minimize the geographical distance between the starting station and the destination station of passenger travel;the second is to aviod traffic congestion and reduce the passenger travel time.The third is to take the number of transfers as the core and minimize the number of transfers in the whole travel process,so as to bring convenience to passengers.According to these research,they have not taken both in-car crowdness and traffic congestion as consideration when they research the passenger riding route recommendation.This paper proposed a passenger riding route recommendation method based on traffic congestion prediction and in-car crowdness prediction.This method includes the traffic congestion prediction model,in-car crowdness prediction model and the riding route recommendation model.Firstly,the time and location information are extracted from GPS trajectory,IC card consumption record,bus station and line vector data,and the bus arrival time and passenger flow at each station are deduced by the map matching and the inference of boarding and alighting stops.The traffic congestion coefficient is estimated by the time interval between the bus travel time and the non-congestion travel time during the peak time period.Meanwhile,the ratio between the number of non-crowded passengers and the total number of boarding passengers in each station is used as the index to measure the crowdness of passengers in the vehicle.Secondly,combined with the weather,time and other spatio-temporal characteristics,according to the historical road congestion and passenger riding crowdness to predict the future data by using LSTM,and introduce the equal interval classification method to classify the road congestion and in-car crowdness.Finally,the score of different routes considering road congestion and in-car crowdness can be calculated by using a weighted function,which have two adjustable weight.According to the score,some scientific routes are recommended and divided into the more time-saving,the more comfortable and the middle in both.In this paper,the performance of this method is tested and evaluated by using the data of six bus lines in Guangzhou,China,the results proves this method is effective and practicable.