Identification Of Taxi Detour Behavior And Its Macro And Micro Feature

The taxi system is an important part of the urban public transport system.As a good complement to the bus network,it is strictly regulated by the transport sector.The GPS(Global Positional System)equipment mounted on the taxi generates a large amount of traffic information-rich taxi movement trajectory data,which has the characteristics of high accuracy,wide coverage and real-time dynamic information.On the one hand,it brings a large amount of available data to the transportation department and related researchers,which enriches the research of taxi system.On the other hand,the massive taxi datasets bring difficulties to the mining,identification and supervision of illegal operation behaviors.Compared to the massive,common,normal data that contains regular patterns,there is a small subset of data with more interesting information,usually containing a small number of abnormal behavioral patterns,associated with some kinds of problem or rare event.Such as the abnormal behavior of taxi detours,which is a long-standing challenge for taxi services.Greedy taxi drivers overcharge passengers by deliberately detouring to extend the driving distance,which is imperceptible to ordinary passengers.At present,the illegal operation of most taxis is mainly detected through manual inspection by experienced staff based on passenger complaints,which is inefficient and time-consuming.In the complex urban road network environment,the existing anomaly detection systems all have problems and shortcomings,such as too single anomaly standard and a high false alarm rate.Therefore,with the advancement of today’s smart law enforcement and off-site law enforcement,an abnormal behavior recognition framework is proposed to accurately identify abnormal orders and taxi detour trajectories.This research is mainly aimed at the illegal operation behavior in the taxi system.Using the taxi order data and trajectory data,a detour behavior detection framework is designed,which realizes the accurate identification of abnormal orders and taxi detour trajectories and analysis the operational characteristics and micro-motion characteristics.In terms of data preprocessing,this study firstly carried out data cleaning work such as coordinate transformation,error data removal,missing attribute completion,etc.,and carried out regional grid and trajectory grid processing respectively,according to the characteristics of the two data sets.Secondly,feature engineering is performed on the two datasets,and the OD of the trajectory data is extracted and grouped for statistics,as the basis for subsequent research.Based on the above results,this study proposes an abnormal detour order detection framework and detour trajectory detection framework from the macro and micro levels.The former is based on taxi order data,using DBSCAN clustering and anomaly detection based on isolation forest.The XGboost classification method and the Shapely are combined to evaluate the feature importance,focus on mining detour behaviors from various abnormal behaviors,and analyze the macro operation characteristics of detour orders.Based on the taxi trajectory data,the latter proposes a detour trajectory anomaly detection technology framework,which combines the iBAT anomaly detection algorithm with the DTW trajectory similarity measurement method to improve the detection rate,and combines the spatiotemporal characteristics to reduce the false detection rate of intentional detour behavior recognition.Finally,this research uses the taxi order data and taxi vehicle positioning information data(trajectory data)generated on the passenger car travel platform in Nanjing to design and verify the experiment.Driving distance and time are considered based on the iBAT algorithm and iBAT+DTW algorithm.By comparing relevant experimental data,the final experimental results show that the method proposed in this study has the lowest misjudgment rate for taxi detour fraud,which is 18% higher than the iBAT algorithm.