Travel Activity Pattern Analysis And Bus Routes Optimization In Large Scale Transit Network

With rapid urbanization and mechanization,the speed of urban transportation infrastructure and road construction in the major cities in our country can not catch up with the increase in the daily travel demands of city residents,leading to growing transportation pressure and environmental pollution.Against this background,the strategy of prioritizing public transportation development has been introduced to boost the planning and construction of urban transit systems,resulting in the increasing expansion of transit networks in cities and promotion of coordinated development across urban regions,as well as alleviating the severe traffic congestion to some extent.However,the growing scale of transit networks in cities has also increased the difficulty in the management and optimization of public transportation systems.On the one hand,it becomes more and more challenging to explore the travel demand characteristics of transit users in large-scale public transportation networks.For example,using traditional analytical tools to estimate transit passengers’ daily travel demands based on manually-collected data is considered impracticable in large-scale transit networks due to high human costs and long investigation time.On the other hand,the expanding transit network also raises the complexity of transit system optimization problems,denying the implementation possibility of many classic optimization methods.In recent years,the advances in computing and information technologies have provided both new opportunities and challenges for addressing the problems facing travel behavior analysis and systems optimization in large-scale transit networks.On one side,technological innovation has stimulated the emergence of automatically-collected transit data and promising data mining and analysis methods.On the other side,these new methods and data can not be directly applied to management and optimization practices in large-scale transit networks without necessary improvements,adaptations and combinations with the domain knowledge of transportation engineering.Based on the practical needs of system management and optimization in large-scale transit networks,this paper focuses on transit passengers’ travel-activity pattern analysis and transit network optimization in large-scale public transportation networks.By merging the state-of-art techniques across multiple disciplines including computer science,information theory and network science,this paper proposes an efficient framework exhibiting remarkable automobility,parallelizability and transferability,to address issues concerning three major topics in the large-scale transit network studies——namely the passenger flow OD estimation,travel-activity pattern identification and analysis,and transit network design.The main purpose of this paper is to foster better understanding of transit users’ travel behavior,optimize transit route sets and eventually improve the operating efficiency of transit systems and reduce passengers’ travel costs.The main contents of this paper contain the following:(1)Estimation of transit passenger flow OD based on automatically-collected data.In the recent decade,existing literature has proposed various trip-chain-based methods for transit O-D estimation using transit smart card data.However,one issue of the classic trip-chaining estimation method could lead to massive data loss: O-D estimation of single trips can not be conducted since the trip-chain approach generally requires at least two trip records per day per passenger to process.To fix this flaw in the conventional trip-chaining approach and fill the research gap,this paper first introduced entropy rate as the travel regularity measurement and an advanced algorithm for estimating entropy rates from passengers’ travel sequences.Then,by mimicking the mechanism of noise reduction in information theory,a minimum entropy-rate model is proposed to infer the alighting of single trips by identifying transit stops that best fit passengers’ travel regularity.A case study based on the Nanjing transit smart card dataset indicates that the proposed method could increase the estimation accuracy by 60% comparing an existing approach called the“muting return hypothesis”(which also concerns the alighting inference of single transit trips).The case study also suggests that the proposed method could improve the estimation success rate by 20% compared to the conventional trip-chain method and obtain an overall estimation accuracy of 80%.Moreover,the relationship between entropy rates and OD estimation accuracy has been investigated based on a sensitivity analysis.(2)Travel-activity pattern identification and analysis for transit passengers.The nonuniformity of existing methods for defining and representing transit passengers’ travel patterns often prevents an in-depth travel pattern analysis in multi-mode transit systems,and limits the type of travel patterns one can identify.In order to address these issues confronting the travel-activity pattern exploration of transit passengers,this paper proposed a new universal method for representing travel patterns for transit riders based on temporal motif ——an emerging concept in network science.Different temporal motifs could be considered as instances of distinct travel-activity patterns.A scalable algorithm is developed to identify temporal motifs from daily trip sub-sequences extracted from two smart card datasets.In the case study of Nanjing metro and bikesharing systems,the proposed method,incorporated with land-use data,exhibits its benefits in revealing the potential correlation between varying topologies of trip combinations and specific activity chains.Commuting,different types of transfer,and other uncommon travel patterns have been recognized.Moreover,different travel-activity patterns,including “Home→Work→Post-work activity(for dining or shopping)→Back home”,“Home→Work→Home→Work”,“Home→Workplace A→Workplace B→Back to workplace A”,have been distinguished.In addition,two application examples of the travel motif have been presented to demonstrate the practicality of the proposed approach.(3)Transit network design based on graph clustering of passenger flows.Transit network optimization problems,such as the Transit Network Design(TND)problem,are generally difficult to be solved within reasonable computational time,especially for large real-world instances.Modifications made in existing TND studies for improving the computational performance could vary across different solution methods,assumptions and problem modeling,raising concerns about their transferability and implementation difficulty.This study presents a general framework to accelerate the TND problem-solving for large instances by decomposing the transit network into smaller subnetworks.We first proposed a modified Louvain algorithm to generate an optimal network partitioning based on large-scale transit OD matrix obtained from OD estimation.A solution method(i.e.,a Genetic Algorithm,GA)is then implemented to address the TND problem in each subnetwork separately and parallelly,thereby saving much computational time.Transit routes generated and optimized in each subnetwork are then combined into one route set as the solution for the entire transit network.The experiment on a well-known benchmark dataset shows that the proposed method,even incorporated with only a basic version of GA,can yield solutions better than most of its existing competitors with an outstanding computational performance.Moreover,the results indicate that our method could speed up the basic GA nearly 1,000 times faster on the iteration process.Another experiment on a real-world network instance(i.e.,the Nanjing bus)shows that our method can optimize the currently operating routes,reducing the average travel time by 18.6% and the number of transfers by 17.7%.(4)Transit network design based on graph clustering of passengers’ travel patterns.In order to design(or optimize)a transit route set more consistent with passengers’ long-established travel behavior and habits,this section proposed a spectral method to partition the large-scale transit networks with the extracted bus passenger travel patterns as model input.Next,GA is introduced to solve each subnetwork’s transit network design problem separately and simultaneously.Benefiting from the main idea of“decomposing the network and optimizing each one”,the proposed approach can yield a considerably good solution for the transit network optimization problem within reasonable computation time while minimizing the potential breaks of passengers’ travel patterns.The mathematical model of the optimization problem and solution method are identical to those of transit network design based on graph clustering of passenger flows for comparison purposes.The case study based on actual ground transit data in Nanjing’s main urban area shows that compared to the original Nanjing bus network case,the route set obtained by solving the transit network optimization problem based on travel-pattern graph clustering resulted in an 8.2% reduction in the average travel time and a 6.8%decrease in the total number of transfers for transit users,while retaining 83.85% of travel-pattern instances of transit passengers,demonstrating its high consistency with passengers’ long-established travel habits.To sum up,this paper has explored and studied three aspects of large-scale bus network research,including travel demand estimation for transit passengers,travel-activity pattern analysis,and transit network design,and proposed a series of new models and algorithms adapted to large-scale bus network research by combining state-of-art methods across multi-disciplines.The research results can provide theoretical and methodological support for better understanding transit passengers’ travel demand characteristics,improving the operational efficiency of large-scale transit systems,and reducing travel costs of transit users.