| Taxi as an important part of urban public transport, with its convenient, fast and other characteristics of service to urban residents. However, since most taxi drivers are not familiar with the distribution of passengers and the road conditions, they often use "cruise" blindly looking for passengers in the no-load driving state, thereby increasing the amount of air travel costs. At the same time, the taxi daily mileage is far more than ordinary private cars, a large number of empty driving mileage on the one hand will cause energy consumption, on the other hand, urban traffic is also a great burden (especially in the peak hours). Based on this problem,this paper builds a recommended model of taxi route based on GPS data, which aims to analyze the surrounding passenger flow rules for taxi drivers and provide some help for their travel decision.This paper first introduces the principle and method of handling the GPS data of the taxi.The SQL database is used to operate and manage the large-scale data. The steps of abnormal data removal and secondary data extraction are introduced in detail. The algorithm of improving the curve fitting is used to matching the GPS points and road network electronic map, lay the foundation for the follow-up study. Then, the index and calculation model which can reflect the operation law of the taxi are established from the aspects of time characteristics and spatial characteristics. At the same time, the concept of potential call hotspot and its passenger probability is put forward when the passenger flow distribution is studied. The calculation model of passenger probability is established, and the potential call hotspots are extracted by density - based OPTICS algorithm.Based on the previous analysis of the data and the study of this issue, the definition and modeling of the taxi passenger recommendation model are put forward, which is to provide the optimal loader sequence and the optimal travel route for the taxi driver, so that the driver travels along the recommended route When the probability of meeting the passenger as large as possible, the expected passenger time as short as possible. In this paper, the optimal route evaluation function S(?) and the optimal loader sequence evaluation function Q(?) are established as the measure of the selection of each sequence and route. In addition, the pruning algorithm is proposed to improve the efficiency of the algorithm. The algorithm can simplify the candidate recommendation set, to a certain extent, speed up the algorithm’s computing speed. At the same time, this paper introduces a method to estimate the waiting time of the taxi, and provides the decision reference for the taxi driver. Based on the previous research,the overall architecture and algorithm of the taxi passenger recommendation model are designed. Finally, the paper extracts a large number of real GPS data of Chengdu taxi and selects the small area to carry on the example analysis, validates the validity of the model and the algorithm, discusses the influence of the parameter setting in the model, and determines the best parameter setting of the model. |
