A modeling method for passengers' connection mode selection behavior under sudden subway disruption

Abstract

The present invention relates to a modeling method for passenger connection mode selection behavior under sudden subway interruption, and belongs to the technical field of passenger emergency connection mode selection behavior. The method obtains the attribute information of sudden subway interruption, determines various types of central stations in the urban subway network, analyzes the explanatory variables that affect the selection of passenger connection modes under sudden subway interruption, calculates the standardized influence weights of various psychological latent variables on the selection behavior of passenger subgroups with various attributes, analyzes the nonlinear relationship between the explanatory variables and the passenger psychology, identifies the key explanatory variables that affect the overall passenger psychology to divide the blocked passenger subgroups, constructs the discrete choice model utility function of the heterogeneous passenger subgroups, and further establishes a connection mode selection behavior model that considers the passenger psychology. The present invention can reveal the impact mechanism of sudden subway interruption events on the psychology of different passenger groups, and realize the reasonable prediction of the passenger connection mode selection behavior after the subway interruption.

Description

Technical Field

[0001] The invention belongs to the technical field of passenger emergency connection mode selection behavior, and in particular relates to a passenger connection mode selection behavior modeling method under sudden subway interruption. Background Art

[0002] With the growth of travel demand and the expansion of network operations, the subway system may face various emergencies in actual operation that may lead to subway service interruptions, such as equipment failures and operational accidents. During the interruption, the subway system may suspend operations, slow down the operation speed, close stations or some lines in the affected area until the problem is resolved, which will have a profound impact on the travel of subway passengers. Therefore, once a sudden subway interruption occurs, the subway operation management department needs to formulate an emergency connection plan in a timely manner according to the actual needs of the blocked passengers, and evacuate the stranded passengers to their destinations efficiently and safely. How to accurately identify the connection needs of the blocked passengers is an important link.

[0003] However, in order to obtain the connection needs of passengers blocked after a sudden subway interruption, it is necessary to focus on analyzing the passengers' connection mode selection behavior. In the case of subway service interruption, the passenger connection mode selection behavior determines what mode of transportation the blocked passengers use to complete their subsequent travel, which reflects the most intuitive connection demand, but it has the properties of being difficult to reproduce, complex and nonlinear, and the influencing factors are complex among different passengers, which makes its measurement always very difficult. Existing studies mostly focus on the connection mode selection behavior of daily residents' commuting, while the behavior research under sudden subway interruption mostly analyzes the subsequent travel plan selection of passengers under the interruption scenario. Its essence is the choice of passenger travel path, which leads to the incomplete study of the connection mode selection behavior under subway interruption events. Therefore, it is necessary to deeply explore the passenger connection mode selection behavior under sudden subway interruption.

[0004] In fact, the key to analyzing the behavior of passengers choosing the connection mode under sudden subway interruption lies in how to analyze the influence mechanism of various psychological latent variables and describe the psychology of blocked passengers. On the one hand, after the sudden subway interruption, the passenger's psychology is first affected by multiple objective factors in the outside world, and the normal travel preference will be broken, which will in turn affect the decision-making of the connection mode. Analyzing the influence mechanism of the passenger's psychological latent variables helps to reveal how these factors act on the passenger's behavioral decision-making. On the other hand, the psychological state and emotional response of blocked passengers will directly affect their choice of connection mode. By describing the psychology of blocked passengers, it is possible to more accurately predict the connection travel mode they may choose, providing a basis for emergency management and service deployment. Therefore, how to reveal the influence mechanism of sudden subway interruption on passenger psychology has become a key problem that needs to be solved in analyzing the behavior of passengers choosing the connection mode, and it has important practical significance. Summary of the invention

[0005] The purpose of the present invention is to solve the problems existing in the background technology and provide a modeling method for passengers' connection mode selection behavior under sudden subway interruption, which can reveal the impact mechanism of sudden subway interruption events on the psychology of different passenger groups and realize the reasonable prediction of passengers' connection mode selection behavior after subway interruption.

[0006] The method of the present invention obtains the attribute information of subway sudden interruption, determines the central stations of various types in the urban subway network, analyzes the explanatory variables that affect the selection of passenger connection mode under subway sudden interruption, selects the psychological latent variables that affect the connection mode selection behavior, fits the influence path between psychological latent variables, calculates the standardized influence weight of each psychological latent variable on the selection behavior of each attribute passenger subgroup, fits the influence path between psychological latent variables and explanatory variables, tests the nonlinear relationship between explanatory variables and passenger psychology, identifies the key explanatory variables that affect the overall passenger psychology to divide the blocked passenger subgroups, constructs the discrete choice model utility function of heterogeneous passenger subgroups, and further establishes the connection mode selection behavior model considering passenger psychology. The present invention can reveal the influence mechanism of subway sudden interruption events on the psychology of different passenger groups, realize the reasonable prediction of passenger connection mode selection behavior after subway interruption, and provide effective technical means for obtaining the connection demand of blocked passengers after subway interruption.

[0007] To achieve the above object, the technical solution of the present invention is: a passenger connection mode selection behavior modeling method under sudden subway interruption, comprising:

[0008] S1. Obtaining subway sudden interruption attribute information;

[0009] S2. Determine the central stations of various types in the urban subway network and analyze the explanatory variables that affect passengers' choice of connection mode under sudden subway interruptions;

[0010] S3. Based on the attribute information of subway sudden interruption, the psychological latent variables that affect the passengers' choice of connection mode are selected, and the standardized influence weights of each psychological latent variable on the choice behavior of each attribute passenger subgroup are calculated;

[0011] S4. Based on the attribute information of subway sudden interruption and the explanatory variables that affect the choice of passenger connection mode under subway sudden interruption, the nonlinear relationship between the explanatory variables and the passenger psychology is analyzed to identify the key explanatory variables that affect the overall passenger psychology;

[0012] S5. Divide the obstructed passenger subgroups based on the standardized influence weights of each psychological latent variable on the selection behavior of each attribute passenger subgroup and the key explanatory variables affecting the overall passenger psychology;

[0013] S6. Based on the nonlinear relationship between the explanatory variables and the passenger psychology, and the divided blocked passenger subgroups, the utility function of the discrete choice model of heterogeneous passenger subgroups is constructed, and a connection mode selection behavior model considering the passenger psychology is established.

[0014] In one embodiment of the present invention, in step S1, the subway sudden interruption attribute information at least includes built environment index information around the subway station, basic attribute information of passengers, travel attribute information of interrupted passengers, psychological attribute information of interrupted passengers and attribute information of subway sudden interruption scene; wherein,

[0015] The built environment indicator information around subway stations shall at least include: the plot ratio, land use mix, road network density, number of bus stops, distance to the nearest bus stop, and distance to the CBD within a radius of 900m around ordinary subway stations, 700m around transfer stations, and 1,200m around terminal stations;

[0016] The basic attribute information of the passenger includes at least: the passenger's socio-economic attribute information and the passenger's travel characteristic attribute information;

[0017] The passenger's socioeconomic attribute information at least includes: the passenger's gender, age, monthly income, occupation, and education level, and numerical values ​​are assigned to each of them; wherein, the gender includes male and female, and is assigned values ​​of 1 and 0 respectively; the age includes at least less than 18 years old, 19-25 years old, 26-45 years old, 46-60 years old, and over 60 years old, and is assigned values ​​of 1, 2, 3, 4, and 5 respectively; the monthly income includes at least less than 2,000 yuan, 2,000-5,000 yuan, 5,000-10,000 yuan, and 1,000-2,000 yuan. 10,000-20,000 yuan, and more than 20,000 yuan, are assigned values ​​of 1, 2, 3, 4, and 5 respectively; the occupations include at least personnel of government agencies and institutions, employees of state-owned enterprises, employees of private enterprises, freelancers, retirees, students, and others, which are assigned values ​​of 1, 2, 3, 4, 5, 6, and 7 respectively; the education level includes at least junior high school and below, high school, junior college and undergraduate, master's degree and above, which are assigned values ​​of 1, 2, 3, and 4 respectively; the above assigned values ​​have no actual meaning and all represent nominal variable codes;

[0018] The passenger travel characteristic attribute information at least includes: passenger travel frequency, whether it has experienced interruption, and information attention, and numerical values ​​are assigned to each of them; wherein, the passenger travel frequency at least includes 1-2 times a day, more than 2 times a day, no more than 5 times a week, no more than 5 times a month, and others, which are assigned values ​​of 1, 2, 3, and 4 respectively; whether it has experienced interruption includes yes and no, which are assigned values ​​of 1 and 0 respectively; the information attention includes whether to check real-time subway information, which are assigned values ​​of 1 and 0 respectively; the above assigned values ​​have no actual meaning and all represent nominal variable codes;

[0019] The interrupted passenger travel attribute information at least includes: the interrupted passenger travel purpose, the interrupted passenger travel distance, the interrupted passenger travel companions, and numerical values ​​are assigned to each of them; wherein, the interrupted passenger travel purpose at least includes going to work, going to school, going out on official business, shopping, cultural tourism, visiting relatives and friends, and others, and the values ​​are assigned to 1, 2, 3, 4, 5, 6, and 7 respectively; the interrupted passenger travel distance at least includes less than 5km, 5-10km, 10-15km, 15-20km, and greater than 20km, and the values ​​are assigned to 1, 2, 3, 4, and 5 respectively; the interrupted passenger travel companions at least include family members, classmates, friends, colleagues, and others, and the values ​​are assigned to 1, 2, 3, 4, and 5 respectively; the above assigned numerical values ​​have no actual meaning and all represent nominal variable codes;

[0020] The psychological attribute information of the interrupted passenger includes at least: the scale data of perceived built environment, the scale data of perceived event intensity, and the scale data of passenger psychological decision-making; the scale data of perceived built environment includes at least: the score of perceived travel convenience scale, the score of perceived travel safety scale, and the score of perceived road accessibility scale; the scale data of perceived event intensity includes at least: the score of event novelty scale, the score of event subversion scale, and the score of event criticality scale; the scale data of passenger psychological decision-making includes at least: the score of behavior attitude scale, the score of subjective norm scale, the score of perceived behavior control scale, the score of behavior intention scale, and the score of actual behavior scale;

[0021] The attribute information of the sudden subway interruption scenario includes at least: the interruption site, interruption duration, interruption occurrence time period, location of the interrupted passengers, and alternative connection methods, and numerical values ​​are assigned to each of them; wherein, the interruption site includes at least various types of central sites in the urban subway network; the interruption duration includes at least less than 15 minutes, 15-30 minutes, 30-45 minutes, 45-60 minutes, and greater than 60 minutes, and are assigned values ​​of 1, 2, 3, 4, and 5 respectively; the interruption occurrence time period includes peak hours and off-peak hours, and are assigned values ​​of 1 and 0 respectively; the location of the interrupted passengers includes already entered the station and not entered the station, and are assigned values ​​of 1 and 0 respectively; the alternative connection methods include at least emergency shuttle buses, regular buses, taxis / online car-hailing vehicles, shared bicycles, and walking, and are assigned values ​​of 1, 2, 3, 4, and 5 respectively; the above-mentioned assigned numerical values ​​have no actual meaning and all represent nominal variable codes.

[0022] In one embodiment of the present invention, the implementation of step S2 is as follows:

[0023] Step S21: for the built environment indicator information around the subway station obtained, the types of urban subway line network sites are divided by using a cluster analysis method, and each cluster center is selected as a central site of each type in the urban subway line network, and each cluster center is selected as an interruption site in the subway sudden interruption scenario attribute information;

[0024] Step S22: for the acquired built environment indicator information around the subway station, basic passenger attribute information, interrupted passenger travel attribute information, interrupted passenger psychological attribute information and subway sudden interruption scenario attribute information, draw radar charts of the distribution results of alternative connection mode selection under different built environment indicators around different subway stations, different basic passenger attributes, different interrupted passenger travel attributes and different subway sudden interruption scenario attributes, and compare the alternative connection modes to obtain the explanatory variables that affect the passenger connection mode selection among different attributes.

[0025] In one embodiment of the present invention, the implementation of step S3 is as follows:

[0026] Step S31: for the psychological attribute information of the interrupted passengers, based on the indicator reliability, combined reliability, convergent validity, and discriminant validity of the verified data by the test factor analysis, select the passenger subgroups traveling by each alternative connection mode according to the basic attribute information of the passengers and the single attribute in the travel attribute information of the interrupted passengers, and obtain the psychological attribute information of each passenger subgroup;

[0027] Step S32, constructing a structural equation model for the psychological attribute information of the passenger subgroups divided in step S31; according to the extended theory of planned behavior, event system theory and stimulus-organism-response theory, set perceived travel convenience, perceived travel safety, perceived road accessibility, event novelty, event subversion, event criticality, perceived behavioral control, behavioral intention and actual behavior as passenger psychological latent variables, set the scale items corresponding to the passenger psychological latent variables as measurement variables, determine the endogenous latent variables and exogenous latent variables of the structural equation model, set the measurement variables corresponding to the endogenous latent variables and the exogenous latent variables as the endogenous explicit variables and the exogenous explicit variables of the structural equation model, and propose hypothetical paths between endogenous latent variables and endogenous latent variables, between endogenous latent variables and exogenous latent variables, between exogenous latent variables and exogenous latent variables, between endogenous latent variables and endogenous explicit variables, and between exogenous latent variables and exogenous explicit variables of passenger subgroups with different attributes under each alternative connection mode;

[0028] Step S33: construct the measurement model and structural model of the structural equation model according to the determined endogenous latent variables, exogenous latent variables, endogenous explicit variables, and exogenous explicit variables; wherein,

[0029] The exogenous explicit variable measurement model is:

[0030]

[0031] In the formula, x is the vector of exogenous explicit variables, or is the exogenous latent variable vector, x is the exogenous factor loading matrix, n x is the exogenous measurement error term, which follows a normal distribution;

[0032] The endogenous explicit variable measurement model is:

[0033]

[0034] In the formula, y is the endogenous explicit variable vector, l is the endogenous latent variable vector, c is the endogenous factor loading matrix, d y is the endogenous measurement error term, which follows a normal distribution;

[0035] The structural model is:

[0036]

[0037] In the formula, ψ is the interaction coefficient matrix between endogenous latent variables, G is the interaction coefficient matrix between internal and external latent variables, i is the result error term, which obeys normal distribution;

[0038] Step S34, combining the measurement model and the structural model constructed in step S33, calculating the standardized path coefficients between the endogenous latent variables and the endogenous manifest variables, and between the exogenous latent variables and the exogenous manifest variables, deleting the endogenous manifest variables and the exogenous manifest variables that failed the test in the hypothesized path constructed in step S32, calculating the chi-square degrees of freedom, approximate root mean square error, comparative fit index, Tucker Lewis index and standardized residual of the structural equation model, and testing the fit of the structural equation model;

[0039] Step S35, according to the structural equation model fitted in step S34, add or delete the hypothesis path that affects the fitting result according to the correction suggestion output by the structural equation model, and obtain the influencing paths between endogenous latent variables and endogenous latent variables, between endogenous latent variables and exogenous latent variables, between exogenous latent variables and exogenous latent variables, between endogenous latent variables and endogenous manifest variables, and between exogenous latent variables and exogenous manifest variables of each attribute passenger subgroup under each alternative connection mode;

[0040] Step S36, constructing a neural network model topology structure according to the structural equation model fitted in step S35; wherein, the exogenous explicit variables are the inputs of the neural network, the exogenous latent variables and the endogenous latent variables are the hidden layer neurons of the neural network, the activation function of each hidden layer neuron is set to be a Sigmoid function, the number of hidden layer layers is determined according to the connection mode between the hidden layer neurons and the number of neurons, the endogenous explicit variables are the outputs of the neural network, the interaction relationship between the exogenous latent variables and the endogenous latent variables is the connection mode between the hidden layer neurons, the weight coefficients between the input and output nodes of the neural network and the neurons correspond to the exogenous / endogenous factor load matrix in the fitted structural equation model, and the weight coefficients between the hidden layer neurons of the neural network correspond to the standardized path coefficients in the fitted structural equation model;

[0041] Step S37, perform model training on the neural network model topology structure constructed in step S36; standardize the input and expected output of the model, initialize the weight coefficients between the nodes of the network, input the values ​​of the exogenous explicit variables in the sample, calculate the output value of each network node step by step according to the neural network model topology structure, the expected output value is the value of the endogenous explicit variable in the sample, calculate the total output error between the model output value and the expected output value, the output error is propagated forward step by step according to the network topology structure, and the Delta rule is used to adjust the weight coefficients between the nodes until the total network error is minimized. The weight coefficient at this time is the standardized influence weight of the psychological latent variables of each attribute passenger subgroup under each alternative connection mode on the selection behavior of each attribute passenger subgroup.

[0042] In one embodiment of the present invention, the implementation of step S4 is as follows:

[0043] Step S41: for the explanatory variables affecting the selection of the passenger connection mode among the different attributes obtained in step S22 and the structural equation model constructed in step S32, the obtained explanatory variables are set as covariates of the structural equation model, and hypothesized paths between the endogenous latent variables and the endogenous latent variables, between the endogenous latent variables and the exogenous latent variables, between the exogenous latent variables and the exogenous latent variables, between the endogenous latent variables and the endogenous manifest variables, between the exogenous latent variables and the exogenous manifest variables, between the endogenous latent variables and the covariates, and between the exogenous latent variables and the covariates are proposed for the passengers' overall psychological state under each alternative connection mode;

[0044] Step S42: construct the measurement model and structural model of the multi-indicator multi-causal model according to the determined endogenous latent variables, exogenous latent variables, endogenous explicit variables, exogenous explicit variables, and covariates. The exogenous explicit variable measurement model is:

[0045]

[0046] In the formula, x is the vector of exogenous explicit variables, or is the exogenous latent variable vector, r is the covariate vector ,x is the exogenous factor loading matrix, α is the coefficient matrix between the exogenous latent variables and the covariates, n x and n ξ is the exogenous measurement error term, which follows a normal distribution;

[0047] The endogenous explicit variable measurement model is:

[0048]

[0049] In the formula, y is the endogenous explicit variable vector, l is the endogenous latent variable vector, c is the endogenous factor loading matrix, β The coefficient matrix between the endogenous latent variables and the covariates, d y and d ξ is the endogenous measurement error term, which follows a normal distribution;

[0050] The structural model is:

[0051]

[0052] In the formula, ψ is the interaction coefficient matrix between endogenous latent variables, G is the interaction coefficient matrix between internal and external latent variables, i is the result error term, which obeys normal distribution;

[0053] Step S43, combine the measurement model and the structural model constructed in step S42, calculate the standardized path coefficients between endogenous latent variables and endogenous manifest variables, between exogenous latent variables and exogenous manifest variables, between endogenous latent variables and covariates, and between exogenous latent variables and covariates, delete the endogenous manifest variables, exogenous manifest variables and covariates that fail the test in the hypothesized path of step S41, calculate the chi-square degrees of freedom, approximate root mean square error, comparative fit index, Tucker Lewis index and standardized residual of the structural equation model, and test the fit of the structural equation model;

[0054] Step S44, according to the structural equation model fitted in step S43, add or delete the hypothesis path that affects the fitting result according to the output of the structural equation model, determine the endogenous latent variables and exogenous latent variables that affect the passenger's connection mode selection behavior, and obtain the influence paths between the endogenous latent variables and the endogenous latent variables, between the endogenous latent variables and the exogenous latent variables, between the exogenous latent variables and the exogenous latent variables, between the endogenous latent variables and the endogenous manifest variables, and between the exogenous latent variables and the exogenous manifest variables of the passenger's overall psychology under each alternative connection mode travel;

[0055] Step S45, constructing a neural network model topology structure according to the structural equation model fitted in step S44; wherein the exogenous explicit variables and covariates are inputs of the neural network, the exogenous latent variables and endogenous latent variables are neurons in the hidden layer of the neural network, the activation function of each hidden layer neuron is set to be a Sigmoid function, the number of hidden layers is determined according to the connection mode between the neurons in the hidden layer and the number of neurons, the endogenous explicit variables are the outputs of the neural network, the interaction relationship between the exogenous latent variables and the endogenous latent variables is the connection mode between the neurons in the hidden layer, the weight coefficients between the input and output nodes of the neural network and the neurons correspond to the exogenous / endogenous factor load matrix in the fitted structural equation model, and the weight coefficients between the neurons in the hidden layer of the neural network correspond to the standardized path coefficients in the fitted structural equation model;

[0056] Step S46, training the model for the neural network model topology structure constructed in step S45; standardizing the input and expected output of the model, initializing the weight coefficients between the nodes of the network, inputting the values ​​of the exogenous explicit variables in the sample, calculating the output values ​​of each network node step by step according to the neural network model topology structure, the expected output value is the value of the endogenous explicit variables in the sample, calculating the total output error between the model output value and the expected output value, the output error is propagated forward step by step according to the network topology structure, and the weight coefficients between the nodes are adjusted using the Delta rule until the total network error is minimized. The weight coefficient at this time is the standardized influence weight of the overall psychological latent variables of the passengers on the overall choice behavior of the passengers under each alternative connection mode;

[0057] Step S47, according to the standardized influence weight calculated in step S46, select the top ten covariates that influence the overall psychological latent variables of passengers under each alternative connection mode, and the corresponding built environment indicators around the subway stations that have the most significant influence on the passengers' psychology, the basic attributes of passengers, the attributes of interrupted passenger travel, and the attributes of sudden subway interruption scenarios, which are the key variables that affect the passengers' psychology.

[0058] In one embodiment of the present invention, the implementation of step S5 is as follows:

[0059] Step S51: Based on the built environment indicators around the subway station with the most significant psychological impact on passengers obtained in step S47, basic attributes of passengers, attributes of interrupted passenger travel, and attributes of subway sudden interruption scenarios, a decision tree with expected classification values ​​for each attribute is obtained by using the CART decision tree algorithm, the classification accuracy of the decision tree training set and the test set is calculated, and a representative subgroup of passengers affected by subway sudden interruptions and their corresponding attributes are determined;

[0060] Step S52: Based on the blocked passenger subgroups and their corresponding attributes determined in step S51, the standardized influence weights of the psychological latent variables of the passenger subgroups of each attribute on the selection behavior of the passenger subgroups of each attribute under each alternative connection mode obtained in step S37 are used to calculate the influence weights of the passenger subgroups according to the combined assignment method. p Select an alternative connection method j The combined weights of the psychological latent variables of the passenger subgroups under each attribute are as follows:

[0061]

[0062] In the formula, W p,i For passenger subgroups p Regarding the combined weight of latent variable i, oh p,q,i For passenger subgroups p In Properties q About latent variables i The weight of .

[0063] In one embodiment of the present invention, step S6 is implemented as follows:

[0064] Step S61: Construct passenger subgroups based on the built environment indicators around the subway station that has the most significant impact on passenger psychology determined in step S47, basic passenger attributes, interrupted passenger travel attributes, subway sudden interruption scene attributes, and the combined weights of passenger subgroups of each attribute calculated in step S52. p Choose from various connection options j The utility function is:

[0065]

[0066]

[0067] In the formula, U p,j For passenger subgroups p Select an alternative connection method j The utility value of V p,j For passenger subgroups p Select an alternative connection method j The utility value of is fixed, e p,j For passenger subgroups p Select an alternative connection method j The utility random error term is m ρ,p,j To influence passenger subgroups p Select an alternative connection method j The covariate weight coefficient is rp,j The influential passenger subgroup obtained in step S51 p Select an alternative connection method j The covariate W η,p,j The influential passenger subgroup obtained in step S52 p Select an alternative connection method j The combined weights of the exogenous latent variables are m η,p,j To influence passenger subgroups p Select an alternative connection method j The exogenous latent variable weight coefficient is or p,j The affected passenger subgroup obtained in step S44 p Select an alternative connection method j The exogenous latent variable, W λ,p,j The influential passenger subgroup obtained in step S52 p Select an alternative connection method j The endogenous latent variable combination weights are m λ,p,j To influence passenger subgroups p Select an alternative connection method j The endogenous latent variable weight coefficient is l p,j The affected passenger subgroup obtained in step S44 p Select an alternative connection method j endogenous latent variables;

[0068] Step S62: Use the utility function constructed in step S61 to construct a connection mode selection behavior model that takes into account the passenger psychology. Then the passenger subgroups p Select an alternative connection method j The probability calculation formula is as follows:

[0069]

[0070] Step S63: Using the passenger-psychological connection mode selection behavior model constructed in step S62, Python Biogeme software is used to complete the parameter m ρ,p,j , m η,p,j , m λ,p,j The calibration was carried out, the model test t value was calculated to eliminate invalid parameters, and the goodness of fit ratio, adjusted goodness of fit ratio, Akaike information criterion AIC, and Bayesian information criterion BIC were calculated to screen and obtain the optimal model results.

[0071] Compared with the prior art, the present invention has the following beneficial effects:

[0072] (1) The present invention introduces the built environment perception scale, event intensity perception scale, and passenger psychological decision scale to design a passenger behavior questionnaire under interruption scenarios, providing technical means and scenario data for comprehensively mapping the passenger connection mode selection behavior under subway sudden interruption scenarios, breaking through the limitation of incomplete interruption scenario data obtained by existing technical solutions;

[0073] (2) Based on the extended theory of planned behavior, the present invention selects key psychological latent variables that affect behavior, and uses a method combining structural equation model and artificial neural network model to analyze the nonlinear relationship between the variables, revealing the impact mechanism of subway disruption events on passenger psychology, providing a more comprehensive behavioral explanation for passenger psychological characterization, and improving the disadvantage of the existing technology that it cannot sort out the nonlinear relationship between the variables that affect choice behavior;

[0074] (3) The present invention combines unobservable psychological latent variables with observable key variables such as personal attribute variables, travel characteristic variables, and scenario variables to jointly construct the utility function of the discrete choice model for different passenger groups, further improving the theoretical research framework of the connection mode selection behavior, and providing a theoretical basis for analyzing the connection mode selection behavior of passengers in emergency situations. It makes up for the deficiency of the existing technology that only models the overall connection mode selection behavior of blocked passengers and it is difficult to screen out characteristic groups for separate modeling. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 is a flow chart of a method for modeling passenger connection mode selection behavior under sudden subway interruption provided by an embodiment of the present invention;

[0076] Figure 2 It is a flowchart of an embodiment of the present invention for determining various types of central stations in an urban subway network and analyzing explanatory variables that affect the choice of passenger connection mode under sudden subway interruption;

[0077] Figure 3 is a flow chart of an embodiment of the present invention for testing the nonlinear relationship between explanatory variables and passenger psychology and identifying key variables that affect passenger psychology;

[0078] Figure 4 is a flow chart of a method for jointly modeling a structural equation model and an artificial neural network constructed in an embodiment of the present invention;

[0079] Figure 5 It is a flow chart of a method for jointly modeling a multi-index multi-causal model and an artificial neural network constructed in an embodiment of the present invention;

[0080] Figure 6 is a flow chart of dividing representative blocked passenger subgroups according to an embodiment of the present invention;

[0081] Figure 7It is a flow chart of a connection mode selection behavior model considering passenger psychology under sudden subway interruption established in an embodiment of the present invention. DETAILED DESCRIPTION

[0082] The technical solution of the present invention is described in detail below in conjunction with the accompanying drawings.

[0083] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0084] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0085] like Figure 1 As shown, the present invention proposes a passenger connection mode selection behavior modeling method under sudden subway interruption, comprising the following steps:

[0086] (1) Obtaining subway sudden interruption attribute information;

[0087] The subway sudden interruption attribute information at least includes built environment index information around the subway station, basic attribute information of passengers, travel attribute information of interrupted passengers, psychological attribute information of interrupted passengers and interruption scene attribute information;

[0088] The built environment indicator data around the subway station shall at least include: the plot ratio, land use mix, road network density, number of bus stations, distance to the nearest bus station, and distance to the CBD within a radius of 900m around ordinary subway stations, 700m around transfer stations, and 1,200m around terminal stations;

[0089] The basic attribute information of the passenger includes at least: the passenger's socio-economic attribute information and the passenger's travel characteristic attribute information;

[0090] The passenger's socioeconomic attribute information includes at least: the passenger's gender, age, monthly income, occupation, and education level;

[0091] Among them, the gender includes male and female, which are assigned values ​​of 1 and 0 respectively; the age includes at least less than 18 years old, 19-25 years old, 26-45 years old, 46-60 years old, and more than 60 years old, which are assigned values ​​of 1, 2, 3, 4, and 5 respectively; the monthly income includes at least less than 2,000 yuan, 2,000-5,000 yuan, 5,000-10,000 yuan, 10,000-20,000 yuan, and more than 20,000 yuan, which are assigned values ​​of 1, 2, 3, 4, and 5 respectively; the occupation includes at least staff of government agencies and institutions, employees of state-owned enterprises, employees of private enterprises, freelancers, retirees, students, and others, which are assigned values ​​of 1, 2, 3, 4, 5, 6, and 7 respectively; the education level includes at least junior high school and below, high school, junior college and undergraduate, master's degree and above, which are assigned values ​​of 1, 2, 3, and 4 respectively; the above assigned values ​​have no actual meaning and all represent nominal variable codes;

[0092] The passenger travel characteristic attribute information includes at least: passenger travel frequency, whether there has been an interruption, and information attention;

[0093] Among them, the passenger travel frequency includes at least 1-2 times a day, more than 2 times a day, no more than 5 times a week, no more than 5 times a month, and others, which are assigned values ​​of 1, 2, 3, and 4 respectively; whether there has been an interruption includes yes and no, which are assigned values ​​of 1 and 0 respectively; the information attention includes whether to check real-time subway information, which are assigned values ​​of 1 and 0 respectively; the above assigned values ​​have no actual meaning and all represent nominal variable codes;

[0094] The interrupted passenger travel characteristic information includes at least: the interrupted passenger travel purpose, the interrupted passenger travel distance, and the interrupted passenger travel companions;

[0095] Among them, the travel purpose of the interrupted passenger includes at least going to work, going to school, going out on official business, shopping, cultural tourism, visiting relatives and friends, and others, which are assigned values ​​of 1, 2, 3, 4, 5, 6, and 7 respectively; the travel distance of the interrupted passenger includes at least less than 5km, 5-10km, 10-15km, 15-20km, and greater than 20km, which are assigned values ​​of 1, 2, 3, 4, and 5 respectively; the travel companions of the interrupted passenger include at least family members, classmates, friends, colleagues, and others, which are assigned values ​​of 1, 2, 3, 4, and 5 respectively; the above assigned values ​​have no actual meaning and all represent nominal variable codes;

[0096] The interruption passenger psychological attribute information at least includes: perception of built environment scale data, event intensity perception scale data, and passenger psychological decision scale data;

[0097] Among them, the data of the perceived built environment scale at least include: the scores of the items of the perceived travel convenience scale, the scores of the items of the perceived travel safety scale, and the scores of the items of the perceived road accessibility scale; the data of the event intensity perception scale at least include: the scores of the items of the event novelty scale, the scores of the items of the event subversiveness scale, and the scores of the items of the event criticality scale; the data of the passenger psychological decision-making scale at least include: the scores of the items of the behavior attitude scale, the scores of the items of the subjective norm scale, the scores of the items of the perceived behavior control scale, the scores of the items of the behavior intention scale, and the scores of the items of the actual behavior scale;

[0098] The subway sudden interruption scenario attribute information at least includes: the interruption site, the interruption duration, the interruption time period, the location of the interrupted passengers, and the alternative connection mode;

[0099] Among them, the interruption site at least includes various types of central sites in the urban subway network; the interruption duration at least includes less than 15 minutes, 15-30 minutes, 30-45 minutes, 45-60 minutes, and more than 60 minutes, which are assigned values ​​of 1, 2, 3, 4, and 5 respectively; the interruption period includes peak and off-peak, which are assigned values ​​of 1 and 0 respectively; the location of the interrupted passengers includes already entered the station and not entered the station, which are assigned values ​​of 1 and 0 respectively; the alternative connection methods at least include emergency shuttle buses, regular buses, taxis / online car-hailing, shared bicycles, and walking, which are assigned values ​​of 1, 2, 3, 4, and 5 respectively; the above-mentioned assigned values ​​have no actual meaning and all represent nominal variable codes;

[0100] (2) Using the acquired information on built environment indicators around subway stations, basic passenger attribute information, travel attribute information of interrupted passengers, psychological attribute information of interrupted passengers, and interruption scenario attribute information, determine the central stations of various types in the urban subway network and analyze the explanatory variables that affect the choice of passenger connection methods under sudden subway interruptions; the flowchart of this step is shown in Figure 2. Figure 2 As shown;

[0101] 1) Based on the built environment indicator data around the subway stations, the cluster analysis method is used to divide the subway line site types, and the cluster centers are selected as the central sites of each type in the urban subway network, and they are selected as the interruption sites in the subway sudden interruption scenario attribute information;

[0102] 2) Based on the built environment indicator data, basic passenger attribute information, interrupted passenger travel attribute information and interruption scenario attribute information around the subway station, draw radar charts of the results of passenger alternative connection mode selection under different built environment indicators around different subway stations, different basic passenger attributes, different interrupted passenger travel attributes and different interruption scenario attributes, and compare the connection modes to obtain the explanatory variables that affect the choice of connection mode in different attributes; the flowchart of this step is shown in the figure below: Figure 3 As shown;

[0103] (3) Based on the psychological attribute information of the interrupted passengers, the psychological latent variables that affect the connection mode selection behavior are selected, and the influence paths between the psychological latent variables are fitted according to the structural equation modeling method. The standardized influence weights of each psychological latent variable on the selection behavior of each attribute passenger subgroup are calculated by combining the artificial neural network model. The flowchart of this step is shown in the figure below: Figure 4 As shown;

[0104] 1) For the psychological attribute information data of the interrupted passengers, the reliability of the indicators, the combined reliability, the convergent validity and the discriminant validity of the data were verified by the test factor analysis. The subgroups of passengers traveling by the alternative connection modes were selected according to the basic attribute information of the passengers and the individual attributes of the interrupted passengers' travel attribute information, and the psychological attribute information data of each passenger subgroup were obtained;

[0105] 2) Construct a structural equation model based on the psychological attribute information of the divided passenger subgroups; according to the extended theory of planned behavior, event system theory and stimulus-organism-response theory, set perceived travel convenience, perceived travel safety, perceived road accessibility, event novelty, event subversion, event criticality, perceived behavioral control, behavioral intention and actual behavior as passenger psychological latent variables, set the scale items corresponding to the passenger psychological latent variables as measurement variables, determine the endogenous latent variables and exogenous latent variables of the structural equation model, and set the corresponding measurement variables of the endogenous latent variables and exogenous latent variables as the endogenous explicit variables and exogenous explicit variables of the structural equation model, and propose hypothetical paths between endogenous latent variables and endogenous latent variables, between endogenous latent variables and exogenous latent variables, between exogenous latent variables and exogenous latent variables, between endogenous latent variables and endogenous explicit variables, and between exogenous latent variables and exogenous explicit variables of passenger subgroups with different attributes under each alternative connection mode;

[0106] 3) Construct the measurement model and structural model of the structural equation model based on the determined endogenous latent variables, exogenous latent variables, endogenous manifest variables, and exogenous manifest variables;

[0107] The exogenous explicit variable measurement model is:

[0108]

[0109] In the formula, x is the vector of exogenous explicit variables, or is the exogenous latent variable vector, x is the exogenous factor loading matrix, n x is the exogenous measurement error term, which follows a normal distribution;

[0110] The endogenous explicit variable measurement model is:

[0111]

[0112] In the formula, y is the endogenous explicit variable vector, l is the endogenous latent variable vector, c is the endogenous factor loading matrix, d y is the endogenous measurement error term, which follows a normal distribution;

[0113] The structural model is:

[0114]

[0115] In the formula, ψ is the interaction coefficient matrix between endogenous latent variables, G is the interaction coefficient matrix between internal and external latent variables, i is the result error term, which obeys normal distribution;

[0116] 4) Combine the constructed measurement model and structural model, calculate the standardized path coefficients between endogenous latent variables and endogenous manifest variables, and between exogenous latent variables and exogenous manifest variables, delete the endogenous and exogenous manifest variables that fail the test in the hypothesized path constructed in step S32, calculate the chi-square degrees of freedom, approximate root mean square error, comparative fit index, Tucker Lewis index and standardized residual of the structural equation model, and test the fit of the structural equation model;

[0117] 5) According to the fitted structural equation model, the output of the structural equation model is used to modify the suggestions for adding or deleting the hypothesized paths that affect the fitting results, and the influencing paths between endogenous latent variables and endogenous latent variables, between endogenous latent variables and exogenous latent variables, between exogenous latent variables and exogenous latent variables, between endogenous latent variables and endogenous manifest variables, and between exogenous latent variables and exogenous manifest variables of each attribute passenger subgroup under each alternative connection mode are obtained;

[0118] 6) Construct the topological structure of the neural network model based on the fitted structural equation model of passenger subgroups with various attributes under each alternative connection mode. Among them, the exogenous explicit variables are the input variables of the neural network, the exogenous latent variables and the endogenous latent variables are the neurons of the hidden layer of the neural network, the activation function of each hidden layer neuron is set to the Sigmoid function, the number of hidden layers is determined according to the connection mode and number of neurons between the hidden layer neurons, the endogenous explicit variables are the outputs of the neural network, the interaction relationship between the exogenous latent variables and the endogenous latent variables is the connection mode between the neurons in the hidden layer, the weight coefficients between the input and output nodes of the neural network and the neurons correspond to the exogenous / endogenous factor load matrix in the fitted structural equation model, and the weight coefficients between the neurons in the hidden layer of the neural network correspond to the standardized path coefficients in the fitted structural equation model;

[0119] 7) Model training is performed for the constructed neural network topology. The input and expected output of the model are standardized, the weight coefficients between the nodes of the network are initialized, the values ​​of the exogenous explicit variables in the input sample are taken, the output values ​​of each network node are calculated step by step according to the neural network topology, the expected output value is the value of the endogenous latent variable in the sample, the total output error between the model output value and the expected output value is calculated, the output error is propagated forward step by step according to the network topology, and the weight coefficients between the nodes are adjusted using the Delta rule until the total network error is minimized. At this time, the weight coefficient is the standardized influence weight of the psychological latent variables of each attribute passenger subgroup on the selection behavior of each attribute passenger subgroup under each alternative connection mode.

[0120] (4) Using the psychological attribute information of interrupted passengers and the explanatory variables that affect passengers' choice of connection mode under sudden subway interruptions, the multi-indicator multi-causal model is used to fit the impact path of each explanatory variable on the overall passenger psychology. The artificial neural network model is used to test the nonlinear relationship between the explanatory variables and passenger psychology, and the key variables that affect passenger psychology are identified. This step is as follows: Figure 5 As shown;

[0121] 1) A multi-indicator and multi-causal model is constructed for the explanatory variables that affect the choice of connection mode among the different attributes obtained and the determined psychological latent variables and manifest variables. The obtained explanatory variables are set as covariates of the structural equation model, and hypothesized paths between the endogenous latent variables and endogenous latent variables, between the endogenous latent variables and exogenous latent variables, between the exogenous latent variables and exogenous latent variables, between the endogenous latent variables and endogenous manifest variables, between the exogenous latent variables and exogenous manifest variables, between the endogenous latent variables and covariates, and between the exogenous latent variables and covariates are proposed for the passengers' overall psychology under each alternative connection mode.

[0122] 2) According to the determined endogenous latent variables, exogenous latent variables, endogenous explicit variables, exogenous explicit variables, and covariates, the measurement model and structural model of the multi-indicator multi-causal model are constructed. The exogenous explicit variable measurement model is:

[0123]

[0124] In the formula, x is the vector of exogenous explicit variables, or is the exogenous latent variable vector, r is the covariate vector ,x is the exogenous factor loading matrix, α is the coefficient matrix between the exogenous latent variables and the covariates, n x and n ξ is the exogenous measurement error term, which follows a normal distribution;

[0125] The endogenous explicit variable measurement model is:

[0126]

[0127] In the formula, y is the endogenous explicit variable vector, l is the endogenous latent variable vector, c is the endogenous factor loading matrix, β The coefficient matrix between the endogenous latent variables and the covariates, d y and d ξ is the endogenous measurement error term, which follows a normal distribution;

[0128] The structural model is:

[0129]

[0130] In the formula, ψ is the interaction coefficient matrix between endogenous latent variables, G is the interaction coefficient matrix between internal and external latent variables, i is the result error term, which obeys normal distribution;

[0131] 3) Combine the constructed measurement model and structural model, calculate the standardized path coefficients between endogenous latent variables and endogenous manifest variables, between exogenous latent variables and exogenous manifest variables, between endogenous latent variables and covariates, and between exogenous latent variables and covariates, delete the endogenous manifest variables, exogenous manifest variables and covariates that failed the test in the hypothesized path of step S41, calculate the chi-square degrees of freedom, approximate root mean square error, comparative fit index, Tucker Lewis index and standardized residual of the structural equation model, and test the fit of the structural equation model;

[0132] 4) According to the fitted structural equation model, add or delete the hypothesis path that affects the fitting results according to the output of the structural equation model, determine the endogenous latent variables and exogenous latent variables that affect the passengers' choice of connection mode, and obtain the influence paths between the endogenous latent variables and the endogenous latent variables, between the endogenous latent variables and the exogenous latent variables, between the exogenous latent variables and the exogenous latent variables, between the endogenous latent variables and the endogenous manifest variables, and between the exogenous latent variables and the exogenous manifest variables of the passengers' overall psychology under each alternative connection mode;

[0133] 5) Construct the topological structure of the neural network model based on the fitted multi-indicator multi-causal model of the overall passengers under each alternative connection mode. Among them, the exogenous explicit variables and covariates are the input of the neural network, the exogenous latent variables and endogenous latent variables are the neurons of the hidden layer of the neural network, the activation function of each hidden layer neuron is set to the Sigmoid function, the number of hidden layers is determined according to the connection mode and number of neurons between the hidden layer neurons, the endogenous explicit variables are the output of the neural network, the interaction relationship between the exogenous latent variables and the endogenous latent variables is the connection mode between the neurons in the hidden layer, the weight coefficients between the input and output nodes of the neural network and the neurons correspond to the exogenous / endogenous factor load matrix in the fitted structural equation model, and the weight coefficients between the neurons in the hidden layer of the neural network correspond to the standardized path coefficients in the fitted structural equation model;

[0134] 6) Perform model training for the constructed neural network topology. Standardize the input and expected output of the model, initialize the weight coefficients between the nodes of the network, input the values ​​of the exogenous explicit variables in the sample, calculate the output values ​​of each network node step by step according to the topology of the neural network model, and the expected output value is the value of the endogenous explicit variable in the sample. Calculate the total output error between the model output value and the expected output value. The output error is propagated forward step by step according to the network topology. Use the Delta rule to adjust the weight coefficients between the nodes until the total network error is minimized. The weight coefficient at this time is the standardized influence weight of the overall psychological latent variables of passengers on the overall choice behavior of passengers under each alternative connection mode.

[0135] 7) According to the calculated standardized impact weights, the top ten covariates that influence the overall psychological latent variables of passengers under each alternative connection mode are selected. The corresponding built environment indicators around the subway stations that have the most significant impact on the passengers' psychology, the basic attributes of passengers, the attributes of interrupted passenger travel, and the attributes of sudden subway interruption scenarios are identified as the key variables that affect the passengers' psychology.

[0136] (5) Using the calculated influence weights and the identified key variables that influence passenger psychology, representative blocked passenger subgroups are divided according to the CART decision tree algorithm; this step is as follows Figure 6 As shown;

[0137] 1) Based on the built environment indicators around the subway stations that have the most significant psychological impact on passengers, basic passenger attributes, interrupted passenger travel attributes, and interruption scene attributes, the CART decision tree algorithm is used to obtain a decision tree with the expected classification of each attribute value, calculate the classification accuracy of the decision tree training set and test set, and determine the representative subgroups of passengers who are blocked by sudden subway interruptions and their corresponding attributes;

[0138] 2) According to the determined blocked passenger subgroups and their corresponding attributes, the standardized influence weights of the psychological latent variables of the passenger subgroups of each attribute on the selection behavior of the passenger subgroups of each attribute under each alternative connection mode are used to calculate the influence of the passenger subgroups according to the combined assignment method. p Select an alternative connection method j The combined weights of the psychological latent variables of the passenger subgroups under each attribute are as follows:

[0139]

[0140] In the formula, W p,i For passenger subgroups p Regarding the combined weight of latent variable i, oh p,q,i For passenger subgroups p In Properties q About latent variables i The weight of .

[0141] (6) Using the nonlinear relationship tested and the different types of blocked passengers, the utility function of the discrete choice model of heterogeneous passenger subgroups is constructed, and a connection mode selection behavior model considering passenger psychology is further established; this step is as follows: Figure 7 As shown;

[0142] 1) Construct passenger subgroups based on the built environment indicators around the subway stations with the most significant psychological impact on passengers, basic passenger attributes, interrupted passenger travel attributes, interrupted scene attributes, and the calculated combined weights of passenger subgroups with each attribute. p Choose from various connection options j The utility function is:

[0143]

[0144]

[0145] In the formula, U p,j For passenger subgroups p Select an alternative connection method j The utility value of V p,j For passenger subgroups p Select an alternative connection method j The utility value of is fixed, e p,j For passenger subgroups p Select an alternative connection method j The utility random error term is m ρ,p,j To influence passenger subgroups p Select an alternative connection methodj The covariate weight coefficient is r p,j The influential passenger subgroup obtained in step S51 p Select an alternative connection method j The covariate W η,p,j The influential passenger subgroup obtained in step S52 p Select an alternative connection method j The combined weights of the exogenous latent variables are m η,p,j To influence passenger subgroups p Select an alternative connection method j The exogenous latent variable weight coefficient is or p,j The affected passenger subgroup obtained in step S44 p Select an alternative connection method j The exogenous latent variable, W λ,p,j The influential passenger subgroup obtained in step S52 p Select an alternative connection method j The endogenous latent variable combination weights are m λ,p,j To influence passenger subgroups p Select an alternative connection method j The endogenous latent variable weight coefficient is l p,j The affected passenger subgroup obtained in step S44 p Select an alternative connection method j endogenous latent variables;

[0146] 2) Using the constructed utility function, a model of the connection mode selection behavior considering the passenger psychology is constructed. Then the passenger subgroups p Select an alternative connection method j The probability calculation formula is as follows:

[0147]

[0148] 3) Using the constructed mixed selection model, Python Biogeme software was selected to complete the parameters m ρ,p,j , m η,p,j , m λ,p,j The calibration was performed, the model test t value was calculated to eliminate invalid parameters, and the goodness of fit ratio, adjusted goodness of fit ratio, Akaike information criterion (AIC), and Bayesian information criterion (BIC) were calculated to screen and obtain the optimal model results.

[0149] In summary, the present invention designs a modeling method for passenger connection mode selection behavior under sudden subway interruption. The modeling method for passenger connection mode selection behavior under sudden subway interruption of the present invention determines various types of central stations in the urban subway network, analyzes the explanatory variables that affect the passenger connection mode selection under sudden subway interruption, calculates the standardized influence weights of each psychological latent variable on the selection behavior of each attribute passenger subgroup, examines the nonlinear relationship between the explanatory variables and the passenger psychology, identifies the key explanatory variables that affect the overall passenger psychology to divide the blocked passenger subgroups, constructs the discrete choice model utility function of different passenger subgroups, and further establishes a connection mode selection behavior model considering the passenger psychology, so as to provide an effective technical means for obtaining the connection needs of blocked passengers after the subway interruption occurs; the design method of the present invention breaks through the limitation of incomplete interruption scenario data obtained by the existing technical solution, improves the disadvantage that the existing technology cannot sort out the nonlinear relationship between the variables affecting the selection behavior, and makes up for the deficiency that the existing technology only models the overall connection mode selection behavior of blocked passengers and is difficult to screen out the characteristic groups for modeling separately.

[0150] The present invention also provides a passenger connection mode selection behavior modeling system under sudden subway interruption, including a memory, a processor, and computer program instructions stored in the memory and capable of being executed by the processor. When the processor executes the computer program instructions, the method steps described above can be implemented.

[0151] The present invention also provides a computer-readable storage medium, on which computer program instructions that can be executed by a processor are stored. When the processor executes the computer program instructions, the method steps described above can be implemented.

[0152] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0153] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure one A process or multiple processes and / or boxes Figure one A device that provides the functions specified in a block or multiple blocks.

[0154] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure one A process or multiple processes and / or boxes Figure one A function specified in one or more boxes.

[0155] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure one A process or multiple processes and / or boxes Figure one A step that specifies a function in one or more boxes.

[0156] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

Claims

1. A passenger connection mode selection behavior modeling method under sudden subway interruption, characterized in that: include: S1. Obtaining subway sudden interruption attribute information; S2. Determine the central stations of various types in the urban subway network and analyze the explanatory variables that affect passengers' choice of connection mode under sudden subway interruptions; S3. Based on the attribute information of subway sudden interruption, the psychological latent variables that affect the passengers' choice of connection mode are selected, and the standardized influence weights of each psychological latent variable on the choice behavior of each attribute passenger subgroup are calculated; S4. Based on the attribute information of subway sudden interruption and the explanatory variables that affect the choice of passenger connection mode under subway sudden interruption, the nonlinear relationship between the explanatory variables and the passenger psychology is analyzed to identify the key explanatory variables that affect the overall passenger psychology; S5. Divide the obstructed passenger subgroups based on the standardized influence weights of each psychological latent variable on the selection behavior of each attribute passenger subgroup and the key explanatory variables affecting the overall passenger psychology; S6. Based on the nonlinear relationship between explanatory variables and passenger psychology and the divided blocked passenger subgroups, the utility function of the discrete choice model of heterogeneous passenger subgroups is constructed, and a connection mode selection behavior model considering passenger psychology is established; In step S1, the subway sudden interruption attribute information at least includes built environment index information around the subway station, basic attribute information of passengers, travel attribute information of interrupted passengers, psychological attribute information of interrupted passengers and attribute information of subway sudden interruption scene; wherein, The built environment indicator information around subway stations shall at least include: the plot ratio, land use mix, road network density, number of bus stops, distance to the nearest bus stop, and distance to the CBD within a radius of 900m around ordinary subway stations, 700m around transfer stations, and 1,200m around terminal stations; The basic passenger attribute information at least includes: passenger socioeconomic attribute information and passenger travel characteristic attribute information; the passenger socioeconomic attribute information at least includes: passenger gender, age, monthly income, occupation, education level, and numerical values ​​are assigned to each of them; the passenger travel characteristic attribute information at least includes: passenger travel frequency, whether it has experienced interruptions, and information attention, and numerical values ​​are assigned to each of them; The interrupted passenger travel attribute information at least includes: the interrupted passenger travel purpose, the interrupted passenger travel distance, and the interrupted passenger travel companions, and numerical values ​​are assigned to each of them; The psychological attribute information of the interrupted passenger includes at least: the scale data of perceived built environment, the scale data of perceived event intensity, and the scale data of passenger psychological decision-making; the scale data of perceived built environment includes at least: the score of perceived travel convenience scale, the score of perceived travel safety scale, and the score of perceived road accessibility scale; the scale data of perceived event intensity includes at least: the score of event novelty scale, the score of event subversion scale, and the score of event criticality scale; the scale data of passenger psychological decision-making includes at least: the score of behavior attitude scale, the score of subjective norm scale, the score of perceived behavior control scale, the score of behavior intention scale, and the score of actual behavior scale; The subway sudden interruption scenario attribute information at least includes: the interruption site, the interruption duration, the interruption time period, the location of the interrupted passengers, and the alternative connection mode, and numerical values ​​are assigned to each of them; The implementation of step S3 is as follows: Step S31: for the psychological attribute information of the interrupted passengers, based on the indicator reliability, combined reliability, convergent validity, and discriminant validity of the verified data by the test factor analysis, select the passenger subgroups traveling by each alternative connection mode according to the basic attribute information of the passengers and the single attribute in the travel attribute information of the interrupted passengers, and obtain the psychological attribute information of each passenger subgroup; Step S32, constructing a structural equation model for the psychological attribute information of the passenger subgroups divided in step S31; according to the extended theory of planned behavior, event system theory and stimulus-organism-response theory, set perceived travel convenience, perceived travel safety, perceived road accessibility, event novelty, event subversion, event criticality, perceived behavioral control, behavioral intention and actual behavior as passenger psychological latent variables, set the scale items corresponding to the passenger psychological latent variables as measurement variables, determine the endogenous latent variables and exogenous latent variables of the structural equation model, set the measurement variables corresponding to the endogenous latent variables and the exogenous latent variables as the endogenous explicit variables and the exogenous explicit variables of the structural equation model, and propose hypothetical paths between endogenous latent variables and endogenous latent variables, between endogenous latent variables and exogenous latent variables, between exogenous latent variables and exogenous latent variables, between endogenous latent variables and endogenous explicit variables, and between exogenous latent variables and exogenous explicit variables of passenger subgroups with different attributes under each alternative connection mode; Step S33: construct the measurement model and structural model of the structural equation model according to the determined endogenous latent variables, exogenous latent variables, endogenous explicit variables, and exogenous explicit variables; wherein, The exogenous explicit variable measurement model is: In the formula, x is the vector of exogenous explicit variables, η is the exogenous latent variable vector, χ is the exogenous factor loading matrix, ν x is the exogenous measurement error term, which follows a normal distribution; The endogenous explicit variable measurement model is: In the formula, y is the endogenous explicit variable vector, λ is the endogenous latent variable vector, γ is the endogenous factor loading matrix, δ y is the endogenous measurement error term, which follows a normal distribution; The structural model is: In the formula, ψ is the interaction coefficient matrix between endogenous latent variables, G is the interaction coefficient matrix between internal and external latent variables, θ is the result error term, which obeys normal distribution; Step S34, combining the measurement model and the structural model constructed in step S33, calculating the standardized path coefficients between the endogenous latent variables and the endogenous manifest variables, and between the exogenous latent variables and the exogenous manifest variables, deleting the endogenous manifest variables and the exogenous manifest variables that failed the test in the hypothesized path constructed in step S32, calculating the chi-square degrees of freedom, approximate root mean square error, comparative fit index, Tucker Lewis index and standardized residual of the structural equation model, and testing the fit of the structural equation model; Step S35, according to the structural equation model fitted in step S34, add or delete the hypothesis path that affects the fitting result according to the correction suggestion output by the structural equation model, and obtain the influencing paths between endogenous latent variables and endogenous latent variables, between endogenous latent variables and exogenous latent variables, between exogenous latent variables and exogenous latent variables, between endogenous latent variables and endogenous manifest variables, and between exogenous latent variables and exogenous manifest variables of each attribute passenger subgroup under each alternative connection mode; Step S36, constructing a neural network model topology structure according to the structural equation model fitted in step S35; wherein, the exogenous explicit variables are the inputs of the neural network, the exogenous latent variables and the endogenous latent variables are the hidden layer neurons of the neural network, the activation function of each hidden layer neuron is set to be a Sigmoid function, the number of hidden layer layers is determined according to the connection mode between the hidden layer neurons and the number of neurons, the endogenous explicit variables are the outputs of the neural network, the interaction relationship between the exogenous latent variables and the endogenous latent variables is the connection mode between the hidden layer neurons, the weight coefficients between the input and output nodes of the neural network and the neurons correspond to the exogenous / endogenous factor load matrix in the fitted structural equation model, and the weight coefficients between the hidden layer neurons of the neural network correspond to the standardized path coefficients in the fitted structural equation model; Step S37, perform model training on the neural network model topology structure constructed in step S36; standardize the input and expected output of the model, initialize the weight coefficients between the nodes of the network, input the values ​​of the exogenous explicit variables in the sample, calculate the output value of each network node step by step according to the neural network model topology structure, the expected output value is the value of the endogenous explicit variable in the sample, calculate the total output error between the model output value and the expected output value, the output error is propagated forward step by step according to the network topology structure, and the Delta rule is used to adjust the weight coefficients between the nodes until the total network error is minimized. The weight coefficient at this time is the standardized influence weight of the psychological latent variables of each attribute passenger subgroup under each alternative connection mode on the selection behavior of each attribute passenger subgroup.

2. According to the method for modeling passenger connection mode selection behavior under sudden subway interruption in claim 1, it is characterized in that: The implementation of step S2 is as follows: Step S21: for the built environment indicator information around the subway station obtained, the types of urban subway line network sites are divided by using a cluster analysis method, and each cluster center is selected as a central site of each type in the urban subway line network, and each cluster center is selected as an interruption site in the subway sudden interruption scenario attribute information; Step S22: for the acquired built environment indicator information around the subway station, basic passenger attribute information, interrupted passenger travel attribute information, interrupted passenger psychological attribute information and subway sudden interruption scenario attribute information, draw radar charts of the distribution results of alternative connection mode selection under different built environment indicators around different subway stations, different basic passenger attributes, different interrupted passenger travel attributes and different subway sudden interruption scenario attributes, and compare the alternative connection modes to obtain the explanatory variables that affect the passenger connection mode selection among different attributes.

3. The passenger connection mode selection behavior modeling method under sudden subway interruption according to claim 1 is characterized in that: The implementation of step S4 is as follows: Step S41: for the explanatory variables affecting the selection of the passenger connection mode among the different attributes obtained in step S22 and the structural equation model constructed in step S32, the obtained explanatory variables are set as covariates of the structural equation model, and hypothesized paths between the endogenous latent variables and the endogenous latent variables, between the endogenous latent variables and the exogenous latent variables, between the exogenous latent variables and the exogenous latent variables, between the endogenous latent variables and the endogenous manifest variables, between the exogenous latent variables and the exogenous manifest variables, between the endogenous latent variables and the covariates, and between the exogenous latent variables and the covariates are proposed for the passengers' overall psychological state under each alternative connection mode; Step S42: construct the measurement model and structural model of the multi-indicator multi-causal model according to the determined endogenous latent variables, exogenous latent variables, endogenous explicit variables, exogenous explicit variables, and covariates. The exogenous explicit variable measurement model is: In the formula, x is the vector of exogenous explicit variables, η is the exogenous latent variable vector, ρ is the covariate vector , χ is the exogenous factor loading matrix, α is the coefficient matrix between the exogenous latent variables and the covariates, ν x and ν ξ is the exogenous measurement error term, which follows a normal distribution; The endogenous explicit variable measurement model is: In the formula, y is the endogenous explicit variable vector, λ is the endogenous latent variable vector, γ is the endogenous factor loading matrix, β The coefficient matrix between endogenous latent variables and covariates, δ y and δ ξ is the endogenous measurement error term, which follows a normal distribution; The structural model is: In the formula, ψ is the interaction coefficient matrix between endogenous latent variables, G is the interaction coefficient matrix between internal and external latent variables, θ is the result error term, which obeys normal distribution; Step S43, combine the measurement model and the structural model constructed in step S42, calculate the standardized path coefficients between endogenous latent variables and endogenous manifest variables, between exogenous latent variables and exogenous manifest variables, between endogenous latent variables and covariates, and between exogenous latent variables and covariates, delete the endogenous manifest variables, exogenous manifest variables and covariates that fail the test in the hypothesized path of step S41, calculate the chi-square degrees of freedom, approximate root mean square error, comparative fit index, Tucker Lewis index and standardized residual of the structural equation model, and test the fit of the structural equation model; Step S44, according to the structural equation model fitted in step S43, add or delete the hypothesis path that affects the fitting result according to the output of the structural equation model, determine the endogenous latent variables and exogenous latent variables that affect the passenger's connection mode selection behavior, and obtain the influence paths between the endogenous latent variables and the endogenous latent variables, between the endogenous latent variables and the exogenous latent variables, between the exogenous latent variables and the exogenous latent variables, between the endogenous latent variables and the endogenous manifest variables, and between the exogenous latent variables and the exogenous manifest variables of the passenger's overall psychology under each alternative connection mode travel; Step S45, constructing a neural network model topology structure according to the structural equation model fitted in step S44; wherein the exogenous explicit variables and covariates are inputs of the neural network, the exogenous latent variables and endogenous latent variables are neurons in the hidden layer of the neural network, the activation function of each hidden layer neuron is set to be a Sigmoid function, the number of hidden layers is determined according to the connection mode between the neurons in the hidden layer and the number of neurons, the endogenous explicit variables are the outputs of the neural network, the interaction relationship between the exogenous latent variables and the endogenous latent variables is the connection mode between the neurons in the hidden layer, the weight coefficients between the input and output nodes of the neural network and the neurons correspond to the exogenous / endogenous factor load matrix in the fitted structural equation model, and the weight coefficients between the neurons in the hidden layer of the neural network correspond to the standardized path coefficients in the fitted structural equation model; Step S46, training the model for the neural network model topology structure constructed in step S45; standardizing the input and expected output of the model, initializing the weight coefficients between the nodes of the network, inputting the values ​​of the exogenous explicit variables in the sample, calculating the output values ​​of each network node step by step according to the neural network model topology structure, the expected output value is the value of the endogenous explicit variables in the sample, calculating the total output error between the model output value and the expected output value, the output error is propagated forward step by step according to the network topology structure, and the weight coefficients between the nodes are adjusted using the Delta rule until the total network error is minimized. The weight coefficient at this time is the standardized influence weight of the overall psychological latent variables of the passengers on the overall choice behavior of the passengers under each alternative connection mode; Step S47: According to the standardized influence weights calculated in step S46, select the top ten covariates that influence the overall psychological latent variables of passengers under each alternative connection mode. The corresponding built environment indicators around the subway stations that have the most significant impact on the passengers' psychology, the basic attributes of passengers, the attributes of interrupted passenger travel, and the attributes of sudden subway interruption scenarios are the key variables that have been identified to affect the passengers' psychology.

4. The passenger connection mode selection behavior modeling method under sudden subway interruption according to claim 3 is characterized in that: The implementation of step S5 is as follows: Step S51: Based on the built environment indicators around the subway station with the most significant psychological impact on passengers obtained in step S47, basic attributes of passengers, attributes of interrupted passenger travel, and attributes of subway sudden interruption scenarios, a decision tree with expected classification values ​​for each attribute is obtained by using the CART decision tree algorithm, the classification accuracy of the decision tree training set and the test set is calculated, and a representative subgroup of passengers affected by subway sudden interruptions and their corresponding attributes are determined; Step S52: Based on the blocked passenger subgroups and their corresponding attributes determined in step S51, the standardized influence weights of the psychological latent variables of the passenger subgroups of each attribute on the selection behavior of the passenger subgroups of each attribute under each alternative connection mode obtained in step S37 are used to calculate the influence weights of the passenger subgroups according to the combined assignment method. p Select an alternative connection method j The combined weights of the psychological latent variables of the passenger subgroups under each attribute are as follows: In the formula, W p,i For passenger subgroups p Regarding the combined weight of latent variable i, ω p,q,i For passenger subgroups p In Properties q About latent variables i The weight of .

5. The method for modeling passenger connection mode selection behavior under sudden subway interruption according to claim 4 is characterized in that: The implementation of step S6 is as follows: Step S61: Construct passenger subgroups based on the built environment indicators around the subway station that has the most significant impact on passenger psychology determined in step S47, basic passenger attributes, interrupted passenger travel attributes, subway sudden interruption scene attributes, and the combined weights of passenger subgroups of each attribute calculated in step S52. p Choose from various connection options j The utility function is: In the formula, U p,j For passenger subgroups p Select an alternative connection method j The utility value of V p,j For passenger subgroups p Select an alternative connection method j The utility value of is fixed, ε p,j For passenger subgroups p Select an alternative connection method j The utility random error term is μ ρ,p,j To influence passenger subgroups p Select an alternative connection method j The covariate weight coefficient is ρ p,j The influential passenger subgroup obtained in step S51 p Select an alternative connection method j The covariate W η,p,j The influential passenger subgroup obtained in step S52 p Select an alternative connection method j The combined weights of the exogenous latent variables are μ η,p,j To influence passenger subgroups p Select an alternative connection method j The exogenous latent variable weight coefficient of η p,j The affected passenger subgroup obtained in step S44 p Select an alternative connection method j The exogenous latent variable, W λ,p,j The influential passenger subgroup obtained in step S52 p Select an alternative connection method j The endogenous latent variable combination weights are μ λ,p,j To influence passenger subgroups p Select an alternative connection method j The endogenous latent variable weight coefficient is λ p,j The affected passenger subgroup obtained in step S44 p Select an alternative connection method j endogenous latent variables; Step S62: Use the utility function constructed in step S61 to construct a connection mode selection behavior model that takes into account the passenger psychology. Then the passenger subgroups p Select an alternative connection method j The probability calculation formula is as follows: Step S63: Using the passenger-psychological connection mode selection behavior model constructed in step S62, Python Biogeme software is used to complete the parameter μ ρ,p,j , μ η,p,j , μ λ,p,j The calibration was carried out, the model test t value was calculated to eliminate invalid parameters, and the goodness of fit ratio, adjusted goodness of fit ratio, Akaike information criterion AIC, and Bayesian information criterion BIC were calculated to screen and obtain the optimal model results.

Images