A reliable identification method of residents' travel mode driven by mobile phone signaling data

Through the Bayesian network model driven by mobile phone signaling data, the time specificity and uncertainty of traffic travel characteristics are extracted, and a reliable travel method identification model is constructed, which solves the accuracy and accuracy of travel method identification in the existing technology, and realizes the realization of travel method. We have a reliable grasp of the structure of urban transportation modes and support for transportation planning.

CN116017407BActive Publication Date: 2024-12-03SOUTHEAST UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202211616958.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-12-15
Publication Date
2024-12-03
Estimated Expiration
2042-12-15

AI Technical Summary

Technical Problem

The existing travel method identification method is difficult to accurately and reliably estimate the multi-mode travel needs and structure, especially when the data accuracy is high, the samples are limited, and the traffic travel characteristics are uncertain, and there is high data acquisition cost and individual activities-traveling. The problem of insufficient characterization of heterogeneity in decision-making behavior.

Method used

Using mobile phone signaling data-driven method, the time specificity and uncertainty of traffic travel characteristics are extracted through the Bayesian network model framework, and a reliable identification model for residents' travel methods is constructed, and the Bayesian network is constructed based on the perspectives of information theory and probability theory. Optimize the model to improve identification accuracy.

Benefits of technology

It has achieved a reliable grasp of the structure of urban transportation modes, reduced the difficulty and cost of data collection, has broader sample coverage and higher identification accuracy, revealed the mechanism of interaction between multiple factors, and supports traffic planning and management decisions. .

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116017407B_ABST
    Figure CN116017407B_ABST
Patent Text Reader

Abstract

The present invention provides a reliable identification method of residents' travel mode driven by mobile phone signaling data, including: obtaining and preprocessing the original data of mobile phone signaling; obtaining residents' travel survey data and matching it with mobile phone signaling data to obtain mobile phone signaling data with travel mode labels; extracting residents' travel characteristics and performing correlation analysis in combination with travel mode labels; constructing a Bayesian network structure based on information theory and a Bayesian network structure based on probability theory to construct a fusion Bayesian network model; processing continuous travel characteristics into an ordered discrete state, completing the learning of Bayesian network model node parameters, and completing the construction of a reliable identification model of residents' travel mode based on the Bayesian network model. The present invention constructs Bayesian networks from the perspectives of information theory and probability theory respectively, takes into account the implicit relationship between travel mode characteristics, uses the BIC function to perform network scoring, and has higher reliability and accuracy in the identification of travel modes.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of travel mode identification, and specifically but not limited to, to a reliable identification method for residents' travel mode driven by mobile phone signaling data. Background Art

[0002] In recent years, with the rapid development of my country's economy and society and the continuous improvement of residents' living standards, consumption concepts, consumption content, and consumption levels have been continuously upgraded, and people's travel needs, travel modes, and residents' travel structure are undergoing profound changes. The continuous growth of the number of motor vehicles and the rise of new green travel modes have affected the multi-modal transportation structure of cities, and have posed new challenges to the identification of urban transportation modes. As an important representation of urban multi-modal transportation demand, residents' travel mode structure data plays a fundamental role in adjusting and optimizing the travel mode structure, building a transportation planning auxiliary decision-making platform, alleviating traffic congestion, and promoting the implementation of transportation emission reduction policies. Limited by the low sample characteristics of traditional residents' travel behavior survey data and the changes in travel characteristics caused by rapid changes in urban spatial structure, the residents' travel division method based on the traditional "four-stage" transportation planning theory method is difficult to accurately and reliably estimate the multi-modal travel demand and structure of the above decision-making applications.

[0003] The existing methods of travel mode identification mainly include: (1) rule-based models, which determine rules through logical features and judge travel feature values. Rules and thresholds are usually set based on the professional knowledge and experience of researchers; (2) unsupervised clustering methods, which aggregate the key features of mobile phone signaling data and classify research samples. Unlabeled data are grouped, and the characteristics of each cluster are analyzed using prior knowledge or combined with other data sources to manually determine the travel mode corresponding to the cluster; (3) analysis methods based on activity theory, based on the aggregate analysis of individual activity travel in the time dimension, considering factors such as time and space constraints, family structure, family background, and personal attributes, to model the mode selection behavior in individual activity-travel decisions; (4) statistical analysis models are divided into multiple models according to whether there is a causal relationship between feature variables. Clear causal relationship features can usually be studied using models such as logit regression, tree structure, and neural network; if the causal relationship is unclear, independence analysis, correlation analysis, and other methods are usually used for research.

[0004] The shortcomings of existing travel mode identification are: (1) The residents' travel mode identification method driven by GPS survey data in existing research does not take into account the problem of data missing and has high requirements for data accuracy. GPS survey data can only extract the travel activity trajectory characteristics of a small number of sample populations, the data skewness characteristics are obvious, and the excavated travel patterns are not universal and representative. (2) The machine learning methods used in existing research cannot effectively reveal the interaction between influencing factors and outcome variables and are unexplainable. (3) The quantitative analysis of traffic travel data in existing research does not take into account the uncertainty of traffic travel characteristics of various travel modes. Travel characteristics are not only related to the inherent properties of transportation tools, but also affected by traffic operation conditions and show time differences. (4) Most of the existing research is based on rule-based models that set rules through logical features, and the threshold selection is highly subjective. (5) Existing research has problems such as insufficient description of the heterogeneity of individual activity-travel decision-making behavior and lack of revelation of the causal mechanism of travel mode selection behavior.

[0005] Therefore, there is an urgent need for a traffic research data foundation and reliable identification methods for traffic modes that are more in line with the characteristics of the times, so as to achieve a reliable grasp of the structure of urban traffic modes and provide effective support for relevant departments in traffic planning, management and decision-making. Faced with the increasingly diverse traffic modes, existing travel mode identification research institutes use traditional manual questionnaire survey data and mobile phone GPS positioning data, which have problems such as high data acquisition costs and limited data samples. The development of mobile communication technology provides a low-cost, large-sample data source for traffic mode identification research. Existing research on the identification of travel modes using mobile phone signaling data mostly uses deterministic models such as rule-based models, machine learning models, and statistical analysis models, which have problems such as insufficient consideration of the uncertainty of traffic travel characteristics and insufficient use of incomplete mobile phone signaling data.

[0006] In view of this, it is necessary to provide a new method to solve at least part of the above problems. Summary of the invention

[0007] In response to one or more problems in the prior art, the present invention proposes a reliable identification method for residents' travel modes driven by mobile phone signaling data. Based on mobile phone signaling data, an effective extraction method for residents' travel characteristics is designed from the perspective of time specificity and uncertainty of traffic travel characteristics. Aiming at the problem of reliable identification of residents' traffic travel modes, a Bayesian network model framework is designed. Through quantitative characterization of the time specificity of traffic travel characteristics, a reliable identification model for residents' traffic travel modes that takes into account the uncertainty of traffic travel characteristics is finally constructed. The present invention can provide effective basic data support for relevant departments to grasp the structure of urban traffic travel modes in the new era, formulate policies for optimizing the structure of traffic travel modes, and promote the development of green transportation and low-carbon transportation.

[0008] The technical solution to achieve the purpose of the present invention is:

[0009] A reliable identification method of residents' travel mode driven by mobile phone signaling data, comprising:

[0010] S1. Obtaining original mobile phone signaling data, and preprocessing the original mobile phone signaling data to obtain preprocessed mobile phone signaling data;

[0011] S2. Acquire resident travel survey data, each piece of resident travel survey data includes departure time, arrival time, departure location, and arrival location, where the departure location and arrival location are represented by traffic community numbers, and feature match the mobile phone signaling data with the resident travel survey data to obtain the mobile phone signaling data with travel mode labels;

[0012] S3, dividing the area to be identified into several traffic sub-areas, extracting residents' travel characteristics based on the time specificity and uncertainty of traffic travel characteristics, wherein the residents' travel characteristics include traffic environment characteristics, individual characteristics of travelers, and traffic travel behavior characteristics, and performing correlation analysis on the residents' travel characteristics in combination with travel mode labels, and quantifying the correlation between traffic travel characteristics with mathematical indicators;

[0013] S4. Design a Bayesian network model framework for reliable identification of residents' travel modes: Based on the correlation between residents' travel characteristics, the causal relationship between residents' travel characteristics is judged in combination with prior knowledge, and the first travel characteristic relationship network is constructed; based on sample data, a machine learning method is used to construct a second travel characteristic relationship network; the BIC function is used to score the model, and the first travel characteristic relationship network and the second travel characteristic relationship network are combined to optimize and construct a Bayesian network model;

[0014] S5. Quantify the uncertainty of the input features of the Bayesian network model, process the continuous travel features into an ordered discrete state, divide the sample data set into a training set and a test set, use the training sample data set to complete the learning of the parameters of each node of the Bayesian network model, evaluate the model accuracy through the test sample data set, and finally complete the construction of a reliable identification model for residents' travel mode based on the Bayesian network model.

[0015] Furthermore, the specific steps of preprocessing in S1 include:

[0016] S1-1, invalid redundant data filtering: filter and remove invalid data and duplicate recorded data in the original data of mobile phone signaling, wherein the invalid data refers to data with missing location area code and cellular cell code;

[0017] S1-2, ping-pong data processing: updating the start time of the first signaling data and the end time of the last signaling data of the ping-pong data to the start time and the end time of the new signaling data, and taking the recording time of the repeated data as the recording time of the new signaling data;

[0018] S1-3. Drift data processing:

[0019] Determine whether the spatial distance between the mobile phone base stations corresponding to the starting cell and the ending cell of each signaling data record exceeds the distance threshold, and divide the spatial distance by the signaling data recording duration to determine whether the switching speed exceeds the switching speed threshold. Signaling data that exceeds the distance threshold and the switching speed threshold are identified as drift data;

[0020] Merge adjacent drift data records: update the start time field of the previous signaling data and the end time field of the next signaling data to the start time and end time field of the new signaling data, and sum the recording duration of the drift data as the recording duration of the new signaling data.

[0021] Furthermore, the specific steps of matching the mobile phone signaling data with the resident travel survey data in S2 include:

[0022] S2-1. According to the gender characteristics of the travelers in each piece of resident travel survey data, the gender characteristics of the users in the mobile phone signaling data are screened, and the mobile phone signaling data with the same gender characteristics as the travelers in the resident travel survey data are retained;

[0023] S2-2. According to the age characteristic attributes of the travelers in each piece of resident travel survey data, the mobile phone signaling data is screened, and the mobile phone signaling data with an age difference of ≤2 years are retained;

[0024] S2-3. Count the number of elements N in the mobile phone signaling data that matches each piece of resident travel survey data. If N=0, that is, the resident travel survey data has no successfully matched mobile phone signaling data, then the resident travel survey data will be removed from the resident travel survey data set; if N=1, that is, the resident travel survey data successfully matches only one mobile phone signaling data, then the resident travel survey data will be retained; if N≥1, that is, there are multiple mobile phone signaling data matching the resident travel survey data, then the resident travel survey data will be removed from the resident travel survey data set.

[0025] Furthermore, the specific steps of S3 include:

[0026] S3-1. Extract traffic environment characteristics and calculate traffic facility index data for each traffic zone, including bus station coverage, bus route duplication, intersection density, road network density and land use mix;

[0027] S3-2, extract individual characteristics of travelers, including gender and age;

[0028] S3-3, extracting travel behavior characteristics, including travel time, travel distance and average travel speed;

[0029] S3-4. Introduce mutual information value to characterize the correlation between the above-mentioned residents' travel characteristics and travel modes. Considering the time-specific influence, calculate the maximum mutual information number during peak hours and off-peak hours respectively and construct the travel characteristic information matrix. The maximum mutual information number calculation formula is as follows:

[0030]

[0031] Among them, MIC(X,Y) represents the maximum mutual information number, x, y are the row elements and column elements in the grid respectively, n X ,n Y Represents the number of grids on the horizontal and vertical axes respectively, and the constraint condition is the number of grids n X n Y <B,(B=n 0.6 ), n represents the total number of samples, p(x,y) is the joint probability density function of the two elements, and p(x), p(y) are the marginal probability density functions of the two elements.

[0032] Furthermore, the specific calculation steps of the bus station coverage index, bus line duplication index, intersection density index, road network density index and land use mix index in S3-1 include:

[0033] 1) Bus stop coverage

[0034]

[0035] Among them, BCR i represents the bus stop coverage rate of transportation community i, S i represents the area of ​​traffic zone i, S j represents the area of ​​a certain radius with bus stop j as the center, n i represents the number of bus stops in transportation zone i, and l represents the number of transportation zones;

[0036] 2) Bus route duplication

[0037]

[0038] Among them, BRRC i represents the duplication of bus routes in transportation community i, L i represents the total length of the road network in traffic zone i, L k represents the length of bus route k, m irepresents the number of bus routes in transportation zone i;

[0039] 3) Road intersection density

[0040] RID i =N / S i ,i=(1,2,...,l)

[0041] Among them, RID i represents the road intersection density of traffic zone i, N represents the number of intersections in traffic zone i, S i represents the area of ​​traffic zone i;

[0042] 4) Road network density

[0043] RND i =L i / S i ,i=(1,2,...,l)

[0044] Among them, RND i represents the road network density of traffic zone i, L i represents the total length of the road network in traffic zone i, S i represents the area of ​​traffic zone i;

[0045] 5) Land use mix

[0046]

[0047] Among them, LM i represents the land use mix of traffic zone i, p q is the percentage of the qth type of land area to the corresponding traffic zone area, Q is the total number of land types, s q is the area of ​​the qth type of land.

[0048] Furthermore, the specific calculation steps of the travel distance and the average travel speed in S3-3 include:

[0049] 1) Travel duration

[0050] The representative timestamp of the starting point of a trip is the departure time, the representative timestamp of the destination is the arrival time, and the difference between the two is the trip duration T I ;

[0051] 2) Travel distance

[0052]

[0053] Among them, D I represents the travel distance of user I, n I represents the total number of intersections that user I passes through, d i,i+1represents the spatial distance between intersection i and intersection i+1;

[0054] 3) Average travel speed

[0055] v I =D I / T I

[0056] Among them, v I represents the average travel speed of user I, T I represents the travel time of user I.

[0057] Furthermore, the specific steps of S4 include:

[0058] S4-1. Based on expert experience and correlation analysis, the dependence or independence of each travel feature in the sense of information theory is determined, and the first Bayesian network model dominated by information theory is constructed;

[0059] S4-2. Based on the hill climbing method, the dependence or independence relationship of each travel feature in the sense of probability theory is obtained, and the second Bayesian network model dominated by probability theory is constructed;

[0060] S4-3. Use BIC measurement as the model structure scoring function to evaluate the structural performance of the above two Bayesian network models. According to the BIC scoring results, based on domain prior knowledge and sample data, the dependence or independence relationship of travel characteristics in the sense of information theory and probability theory is integrated to finally construct the Bayesian network topology structure.

[0061] Furthermore, the specific steps of S5 include:

[0062] S5-1. Characterization of uncertainty in travel characteristics using Bayesian networks: Statistical analysis is performed on sample data of different travel modes. Based on prior knowledge in the field of transportation, the uncertainty of travel characteristics is expressed in the form of probability, forming a physical characterization scheme for uncertainty, and discretizing continuous characteristic variables into ordered discrete states. Resident travel characteristics include traveler gender, traveler age, travel duration, travel distance, average travel speed, public transportation system service level, road system construction level, and land use level.

[0063] S5-2. Divide the mobile phone signaling data with travel mode labels into two data sets, 80% of which is used as a test data set for the node parameters of the Bayesian network model, and 20% is used as a verification data set for the subsequent model effect. Based on the test data set, the node parameters of the Bayesian network model are learned to obtain the conditional probability table of each node in the Bayesian network model, and the travel mode identification model is constructed.

[0064] Furthermore, the formula for learning the node parameters of the Bayesian network model is as follows:

[0065]

[0066] Among them, θ refers to the maximum posterior probability of θ, D represents the sample data set, and P(D) is the prior distribution of the sample data set.

[0067] Compared with the prior art, the present invention adopts the above technical solution and has the following technical effects:

[0068] 1. The reliable identification method of residents' travel mode driven by mobile phone signaling data of the present invention uses mobile phone signaling data, which can reduce the difficulty and cost of data collection compared to GPS data, and has a wider sample coverage.

[0069] 2. The reliable identification method of residents' travel modes driven by mobile phone signaling data of the present invention takes into account that the travel characteristics of various types of transportation modes are jointly affected by multiple uncertain factors in the transportation system, such as the attributes of travelers, transportation facility conditions, and transportation operating environment. It can reveal the mechanism of the mutual influence of multiple factors and clarify the typical scenarios of travel characteristics of various types of travel modes.

[0070] 3. The reliable identification method of residents' travel modes driven by mobile phone signaling data of the present invention constructs Bayesian networks from the perspectives of information theory and probability theory, takes into account the implicit relationship between travel mode characteristics, and uses the BIC function for network scoring, which has higher reliability and accuracy in the identification of travel modes. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] The accompanying drawings are used to provide a further understanding of the present invention and are used together with the description to explain the embodiments of the present invention, but do not constitute a limitation of the present invention. In the accompanying drawings:

[0072] Figure 1 A schematic diagram of travel duration calculation in the reliable identification method of residents' travel mode driven by mobile phone signaling data of the present invention is shown.

[0073] Figure 2 The topological structure diagram of the Bayesian network constructed based on information theory in the reliable identification method of residents' travel mode driven by mobile phone signaling data of the present invention is shown.

[0074] Figure 3 The topological structure diagram of the Bayesian network constructed based on probability theory in the reliable identification method of residents' travel mode driven by mobile phone signaling data of the present invention is shown.

[0075] Figure 4 The result diagram of the mutual information matrix of travel characteristics during peak hours in the reliable identification method of residents' travel modes driven by mobile phone signaling data of the present invention is shown.

[0076] Figure 5The result diagram of the mutual information matrix of travel characteristics during off-peak hours in the reliable identification method of residents' travel modes driven by mobile phone signaling data of the present invention is shown.

[0077] Figure 6 The topological structure diagram of the Bayesian network constructed by integrating information theory and probability theory in the reliable identification method of residents' travel mode driven by mobile phone signaling data of the present invention is shown.

[0078] Figure 7 The distribution histogram of the travel mode selection ratio of travelers of different genders in the reliable identification method of residents' travel mode driven by mobile phone signaling data of the present invention is shown.

[0079] Figure 8 The histogram of the gender ratio distribution of travelers with different travel modes in the reliable identification method of residents' travel modes driven by mobile phone signaling data of the present invention is shown.

[0080] Fig. 9 The histogram of the distribution ratio of travel mode selection of travelers of different ages in the reliable identification method of residents' travel mode driven by mobile phone signaling data of the present invention is shown.

[0081] Fig.10 The distribution probability density diagram of travel time of different travel modes in the reliable identification method of residents' travel modes driven by mobile phone signaling data of the present invention is shown.

[0082] Fig.11 The travel distance distribution probability density diagram of different travel modes in the reliable identification method of residents' travel modes driven by mobile phone signaling data of the present invention is shown.

[0083] Fig.12 The probability density diagram of the average speed distribution of different travel modes in the reliable identification method of residents' travel modes driven by mobile phone signaling data of the present invention is shown.

[0084] Fig.13 The distribution probability density diagram of bus stop coverage rate of travel communities with different travel modes in the reliable identification method of residents' travel modes driven by mobile phone signaling data of the present invention is shown.

[0085] Fig.14 The probability density diagram of the distribution of repetition coefficients of bus routes in travel communities with different travel modes in the reliable identification method of residents' travel modes driven by mobile phone signaling data of the present invention is shown.

[0086] Fig.15 The probability density diagram of the intersection density distribution of travel communities with different travel modes in the reliable identification method of residents' travel modes driven by mobile phone signaling data of the present invention is shown.

[0087] Fig.16The probability density diagram of the road network density distribution of different travel modes in the reliable identification method of residents' travel mode driven by mobile phone signaling data of the present invention is shown.

[0088] Fig.17 The probability density diagram of the mixed degree distribution of travel area land use for different travel modes in the reliable identification method of residents' travel mode driven by mobile phone signaling data of the present invention is shown.

[0089] Fig.18 The figure shows the ratio of the real travel mode structure and the inferred travel mode structure of the verification data set in the reliable identification method of residents' travel mode driven by mobile phone signaling data of the present invention.

[0090] Fig.19 A flow chart of the method for reliably identifying residents' travel modes driven by mobile phone signaling data of the present invention is shown. DETAILED DESCRIPTION

[0091] In order to further understand the present invention, preferred embodiments of the present invention are described below in conjunction with examples. However, it should be understood that these descriptions are only for further illustrating the features and advantages of the present invention, rather than limiting the claims of the present invention.

[0092] The description in this section is only for typical embodiments, and the present invention is not limited to the scope of the embodiments. The combination of different embodiments, the mutual replacement of some technical features in different embodiments, and the mutual replacement of the same or similar prior art means with some technical features in the embodiments are also within the scope of the description and protection of the present invention.

[0093] A reliable identification method of residents' travel mode driven by mobile phone signaling data, comprising:

[0094] S1. Obtaining original mobile phone signaling data, and preprocessing the original mobile phone signaling data to obtain preprocessed mobile phone signaling data;

[0095] S2. Acquire resident travel survey data, each piece of resident travel survey data includes departure time, arrival time, departure location, and arrival location, where the departure location and arrival location are represented by traffic community numbers, and feature match the mobile phone signaling data with the resident travel survey data to obtain the mobile phone signaling data with travel mode labels;

[0096] S3, dividing the area to be identified into several traffic sub-areas, extracting residents' travel characteristics based on the time specificity and uncertainty of traffic travel characteristics, wherein the residents' travel characteristics include traffic environment characteristics, individual characteristics of travelers, and traffic travel behavior characteristics, and performing correlation analysis on the residents' travel characteristics in combination with travel mode labels, and quantifying the correlation between traffic travel characteristics with mathematical indicators;

[0097] S4. Design a Bayesian network model framework for reliable identification of residents' travel modes: Based on the correlation between residents' travel characteristics, the causal relationship between residents' travel characteristics is judged in combination with prior knowledge, and the first travel characteristic relationship network is constructed; based on sample data, a machine learning method is used to construct a second travel characteristic relationship network; the BIC function is used to score the model, and the first travel characteristic relationship network and the second travel characteristic relationship network are combined to optimize and construct a Bayesian network model;

[0098] S5. Quantify the uncertainty of the input features of the Bayesian network model, process the continuous travel features into an ordered discrete state, divide the sample data set into a training set and a test set, use the training sample data set to complete the learning of the parameters of each node of the Bayesian network model, evaluate the model accuracy through the test sample data set, and finally complete the construction of a reliable identification model for residents' travel mode based on the Bayesian network model.

[0099] Aiming at the uncertainty of the travel characteristics of various travel modes under different operating conditions of the traffic system, the present invention analyzes the correlation between the travel characteristics during peak hours and off-peak hours, and proposes a travel speed variation coefficient index to evaluate the influence of the travel characteristics of different travel modes on the travel characteristics of different travel modes. At the same time, aiming at the uncertainty of multi-dimensional travel characteristics, a physical characterization scheme of uncertainty is proposed, and the dependency between travel characteristics is studied from the perspective of information theory and probability theory. By using the Bayesian network structure learning and parameter learning method, a reliable identification model of residents' travel mode based on Bayesian network is constructed.

[0100] The present invention aims at the uncertainty of the travel characteristics of various travel modes under different operating conditions of the transportation system, conducts correlation analysis on the travel characteristics during peak hours and off-peak hours, and proposes a travel speed variation coefficient index to evaluate the influence of the travel characteristics of different travel modes on the attributes of the travel period. It takes into account the common influence of various uncertain factors such as the attributes of travelers, transportation facilities, and transportation operation environment in the transportation system on the travel characteristics of various travel modes, which can reveal the mechanism of the mutual influence of multiple factors and clarify the typical scenarios of the travel characteristics of various travel modes. At the same time, in view of the uncertainty of multi-dimensional transportation travel characteristics, an uncertainty physical representation scheme is proposed, and the dependency relationship between travel characteristics is studied from the perspective of information theory and probability theory. The Bayesian network structure learning and parameter learning method is used to consider the implicit relationship between travel mode characteristics, and the BIC function is used for network scoring, realizing the construction of a reliable identification model of residents' travel modes based on the Bayesian network, which has higher reliability and accuracy in the identification of travel modes.

[0101] Mobile phone signaling data refers to a series of control instructions generated by the mobile communication network to actively or passively, regularly or irregularly maintain contact with the mobile terminal of the mobile phone user, including fields such as mobile phone identification code, timestamp, event type, base station number, base station latitude and longitude, number location, etc. Compared with GPS data, mobile phone signaling data can reduce the difficulty and cost of data collection and has a wider sample coverage.

[0102] Example 1

[0103] A reliable identification method of residents' travel mode driven by mobile phone signaling data includes the following steps:

[0104] In step S1, the original data is mainly 4G mobile phone signaling data provided by China Mobile Communications Company in Kunshan City, Jiangsu Province, and also includes some 2G and 3G mobile phone signaling data. Such data all use COO positioning technology, and the positioning accuracy depends on the service range and installation density of the base station. Table 1 lists the field information of some mobile phone base stations in Kunshan City, mainly including location area code, base station cell code, base station name, and base station latitude and longitude.

[0105] Table 1 Kunshan base station information table

[0106]

[0107]

[0108] The data fields and meanings of the 4G signaling data of the research sample are shown in Table 2, which mainly include information such as the location area code triggered by the signaling data, the base station cell code, and the signaling triggering time.

[0109] Table 2 Meaning of mobile phone signaling data fields

[0110]

[0111] In step S2, the personal characteristics of the travelers and mobile phone users are combined to complete the final feature matching of the residents' travel survey data and the mobile phone signaling data. Among them, type is the type of travel mode, 1 is walking, 2 is bicycle, 3 is car, 4 is bus, and 5 is car. The mobile phone signaling data with travel tags is shown in Table 3 below.

[0112] Table 3. Examples of some matching data

[0113]

[0114]

[0115] In step S3, based on the travel mode labels and three types of travel characteristics corresponding to the mobile phone signaling data, the correlation is tested by calculating the maximum mutual information coefficient of every two variables. The variable information is shown in Table 4 below.

[0116] Table 4 Traffic travel feature fields

[0117]

[0118] The maximum mutual information coefficient matrix is ​​as follows Figure 5-Figure 6 As shown, the horizontal and vertical axes correspond to the variable numbers in Table 4, and the values ​​of the maximum mutual information are all between 0 and 1. The larger the value, the higher the degree of dependence. Figure 5 and Figure 6 It can be seen that the travel mode has a low degree of dependence on individual attribute characteristics, a high degree of dependence on traffic travel behavior characteristics, and the lowest degree of dependence on traffic environment characteristics including bus system service level, road system construction level, and land use level. Whether in peak hours or off-peak hours, the two variables of travel distance and average travel speed are highly correlated with the travel mode, indicating that these two characteristics play a key role in identifying the travel mode; among the traffic environment characteristics, the coverage rate of bus stops in the traffic community has the strongest correlation with the travel mode, and other traffic environment characteristics are related to it.

[0119] In step S4, the structural performance of the two Bayesian networks is compared using the BIC measure as the model structure scoring function. The results are shown in Table 5:

[0120] Table 5 Bayesian network topology structure BIC score table

[0121]

[0122] From the BIC score results, the Bayesian network structure built based on information theory is better than the Bayesian network structure learned from data based on probability theory. This is because the model structure complexity learned by the hill climbing algorithm is significantly higher than that of the self-built network, which will increase the computational cost of the Bayesian network. Therefore, the optimized Bayesian network structure is as follows Figure 6 shown.

[0123] In step S5, different modes of transportation in the transportation system will present different travel characteristics; and under the influence of factors such as transportation facilities and traffic operation conditions, the same mode of transportation will also show different travel characteristics. The uncertainty of travel characteristics is expressed in the form of probability to form an uncertainty physical representation scheme, and the continuous characteristic variables are discretized into some ordered discrete states.

[0124] 1) Gender of travelers

[0125] The travel mode preferences of travelers of different genders and the distribution of travelers of different genders in the same travel mode are analyzed from a statistical perspective. Figure 7 and Figure 8 shown.

[0126] 2) Traveler age

[0127] Statistical analysis of the age distribution of sample data shows that the age distribution of travelers with different travel modes presents different waveforms, such as Fig. 9 shown.

[0128] 3) Travel duration

[0129] By fitting the travel time of different travel modes, we can get the distribution probability density curve of the travel time of Kunshan residents in different travel modes, such as Fig.10 shown.

[0130] 4) Travel distance

[0131] By fitting the travel distances of different travel modes, we can get the distribution probability density curves of the travel distances of Kunshan residents by different travel modes, such as Fig.11 shown.

[0132] 5) Average travel speed

[0133] By fitting the average travel speed of different travel modes, we can get the probability density curve of the average travel speed of Kunshan residents in different travel modes, such as Fig.12 shown.

[0134] 6) Public transportation system service level

[0135] According to different travel modes, the probability density function curves of bus stop coverage and bus line repetition coefficient distribution of the starting and ending communities are fitted respectively, as shown in the figure. Fig.13 , Fig.14 shown.

[0136] 7) Road system construction level

[0137] According to different travel modes, the probability density function curves of the road intersection density and road network density distribution of the starting point and end points of the trip are fitted respectively, such as Fig.15 , Fig.16 shown.

[0138] 8) Land use level

[0139] According to different travel modes, the probability density function curves of the mixed degree distribution of the starting and ending areas of the trip are fitted, such as Fig.17 shown.

[0140] All characteristic variables are discretized into random variables with finite states. Table 6 lists all discretized travel characteristic variables and the meaning of their discrete states.

[0141] Table 6 Discrete states of travel characteristic variables

[0142]

[0143]

[0144] The mobile phone signaling data with travel mode labels are divided into data sets, 80% of which are used as the data set for learning the node parameters of the Bayesian network model, and the remaining 20% ​​will be used for the subsequent model effect test. The quantity distribution of each travel mode is shown in Table 7.

[0145] Table 7 Distribution of the number of training sets and test sets

[0146]

[0147] Based on the test set, the model node parameters are learned to obtain the conditional probability table of each node in the Bayesian network. The average travel speed is taken as an example, as shown in Table 8.

[0148] Table 8. Node conditional probability table for average speed of trip

[0149]

[0150] By analyzing the conditional probability distribution of the average speed nodes of the trip, it can be seen that the speed of walking is significantly lower than that of other modes, and there is a certain overlap between the upper end of the speed range and the speed range of bicycle travel; the speed distribution ranges of electric bicycles and public transportation are very similar, and the overall speed of electric bicycles is slightly higher than that of buses. Other feature nodes are needed to assist in identifying these two modes; the speed distribution range of car travel is significantly higher than that of other travel modes, but there is a certain overlap with the speed of public transportation and electric bicycles. Other feature nodes are also needed to assist in identifying the travel mode.

[0151] Based on the set of observations of travel characteristics in the validation data set, the probability of belonging to each travel mode can be inferred. The travel mode with the highest probability is taken as the Bayesian network inference result. The recognition accuracy of the model for the Kunshan mobile phone signaling validation data set is 82.91%. The classification confusion matrix of the validation data set is shown in Table 9.

[0152] Table 9. Classification confusion matrix of validation dataset

[0153]

[0154] From an overall perspective, the differences between the travel mode structure identified by the model and the actual travel structure are compared and analyzed. The results are as follows: Fig.18 As can be seen from the figure, the Bayesian network has a good grasp of the structure of travel modes, especially the structural proportions of walking, electric bicycles and cars are very close to the actual proportions.

[0155] Based on mobile phone signaling data, the present invention conducts in-depth analysis from the perspective of time specificity and uncertainty of traffic travel characteristics, and designs an effective method for extracting residents' travel characteristics. Aiming at the problem of reliable identification of residents' transportation travel modes, a Bayesian network model framework is designed, and then through the quantitative characterization of the time specificity of traffic travel characteristics, a reliable identification model of residents' transportation travel modes that takes into account the uncertainty of traffic travel characteristics is finally constructed. This method can provide effective basic data support for relevant departments to grasp the structure of urban transportation travel modes in the new era, formulate policies to optimize the structure of transportation travel modes, and promote the development of green transportation and low-carbon transportation.

[0156] The description and application of the present invention here are illustrative, and it is not intended to limit the scope of the present invention to the above-mentioned embodiments. The effects or advantages involved in the specification may not be reflected in the actual experimental examples due to the uncertainty of specific condition parameters or other factors, and the effects or advantages are not used to limit the scope of the invention. The deformation and change of the embodiments disclosed here are possible, and the replacement of the embodiments and the various equivalent parts are well known to those of ordinary skill in the art. It should be clear to those skilled in the art that the present invention can be implemented in other forms, structures, arrangements, proportions, and with other components, materials and parts without departing from the spirit or essential features of the present invention. Other deformations and changes can be made to the embodiments disclosed here without departing from the scope and spirit of the present invention.

Claims

1. A reliable identification method of residents' travel mode driven by mobile phone signaling data, characterized in that: include: S1. Obtaining original mobile phone signaling data, and preprocessing the original mobile phone signaling data to obtain preprocessed mobile phone signaling data; wherein the specific steps of preprocessing include: S1-1, invalid redundant data filtering: filter and remove invalid data and duplicate recorded data in the original data of mobile phone signaling, wherein the invalid data refers to data with missing location area code and cellular cell code; S1-2, ping-pong data processing: updating the start time of the first signaling data and the end time of the last signaling data of the ping-pong data to the start time and the end time of the new signaling data, and taking the recording time of the repeated data as the recording time of the new signaling data; S1-3. Drift data processing: Determine whether the spatial distance between the mobile phone base stations corresponding to the starting cell and the ending cell of each signaling data record exceeds the distance threshold, and divide the spatial distance by the signaling data recording duration to determine whether the switching speed exceeds the switching speed threshold. Signaling data that exceeds the distance threshold and the switching speed threshold are identified as drift data; Merge adjacent drift data records: update the start time field of the previous signaling data and the end time field of the next signaling data to the start time and end time field of the new signaling data, and sum the record duration of the drift data as the record duration of the new signaling data; S2. Acquire resident travel survey data, where each piece of resident travel survey data includes departure time, arrival time, departure location, and arrival location, where the departure location and arrival location are represented by traffic community numbers, and feature match the mobile phone signaling data with the resident travel survey data to obtain mobile phone signaling data with travel mode labels; wherein the specific steps of feature matching include: S2-1. According to the gender characteristics of the travelers in each piece of resident travel survey data, the gender characteristics of the users in the mobile phone signaling data are screened, and the mobile phone signaling data with the same gender characteristics as the travelers in the resident travel survey data are retained; S2-2. According to the age characteristic attributes of the travelers in each piece of resident travel survey data, the mobile phone signaling data is screened, and the mobile phone signaling data with an age difference of ≤2 years are retained; S2-3. Count the number of elements N in the mobile phone signaling data matched by each piece of resident travel survey data. If N=0, that is, the piece of resident travel survey data has no successfully matched mobile phone signaling data, then the piece of resident travel survey data is removed from the resident travel survey data set; if N=1, that is, the piece of resident travel survey data successfully matches only one piece of mobile phone signaling data, then the piece of resident travel survey data is retained; if N≥1, that is, there are multiple mobile phone signaling data matching the piece of resident travel survey data, then the piece of resident travel survey data is removed from the resident travel survey data set; S3. Divide the area to be identified into several traffic zones, extract residents' travel characteristics based on the time specificity and uncertainty of traffic travel characteristics, the residents' travel characteristics include traffic environment characteristics, individual characteristics of travelers, and traffic travel behavior characteristics, and conduct correlation analysis on residents' travel characteristics in combination with travel mode labels, and quantify the correlation between traffic travel characteristics with mathematical indicators; the specific steps include: S3-1. Extract traffic environment characteristics and calculate traffic facility index data for each traffic zone, including bus station coverage, bus route duplication, intersection density, road network density and land use mix; S3-2, extract individual characteristics of travelers, including gender and age; S3-3, extracting travel behavior characteristics, including travel time, travel distance and average travel speed; S3-4. Introduce mutual information value to characterize the correlation between the above-mentioned residents' travel characteristics and travel modes. Considering the time-specific influence, calculate the maximum mutual information number during peak hours and off-peak hours respectively and construct the travel characteristic information matrix. The maximum mutual information number calculation formula is as follows: Among them, MIC(X,Y) represents the maximum mutual information number, x, y are the row elements and column elements in the grid respectively, n X ,n Y Represents the number of grids on the horizontal and vertical axes respectively, and the constraint condition is the number of grids |n X ||n Y |<B,B=n 0.6 , n represents the total number of samples, p(x,y) is the joint probability density function of the two elements, p(x), p(y) are the marginal probability density functions of the two elements; S4. Design a Bayesian network model framework for reliable identification of residents' travel modes: Based on the correlation between residents' travel characteristics, the causal relationship between residents' travel characteristics is judged in combination with prior knowledge, and the first travel characteristic relationship network is constructed; based on sample data, a machine learning method is used to construct a second travel characteristic relationship network; the BIC function is used to score the model, and the first travel characteristic relationship network and the second travel characteristic relationship network are combined to optimize and construct a Bayesian network model; the specific steps include: S4-1. Based on expert experience and correlation analysis, the dependence or independence of each travel feature in the sense of information theory is determined, and the first Bayesian network model dominated by information theory is constructed; S4-2. Based on the hill climbing method, the dependence or independence relationship of each travel feature in the sense of probability theory is obtained, and the second Bayesian network model dominated by probability theory is constructed; S4-3. Use BIC measurement as the model structure scoring function to evaluate the structural performance of the two Bayesian network models. According to the BIC scoring results, based on domain prior knowledge and sample data, the dependence or independence relationship of travel characteristics in the sense of information theory and probability theory is integrated to finally construct the Bayesian network topology structure. S5. Quantify the uncertainty of the input features of the Bayesian network model, process the continuous travel features into an ordered discrete state, divide the sample data set into a training set and a test set, use the training sample data set to complete the learning of the parameters of each node of the Bayesian network model, evaluate the model accuracy through the test sample data set, and finally complete the construction of a reliable identification model for residents' travel mode based on the Bayesian network model.

2. The reliable identification method of residents' travel mode driven by mobile phone signaling data according to claim 1 is characterized in that: The specific calculation steps of bus station coverage index, bus line duplication index, intersection density index, road network density index and land use mix index in S3-1 include: 1) Bus stop coverage Among them, BCR i represents the bus stop coverage rate of transportation community i, S i represents the area of ​​traffic zone i, S j represents the area of ​​a certain radius with bus stop j as the center, n i represents the number of bus stops in transportation zone i, and l represents the number of transportation zones; 2) Bus route duplication Among them, BRRC i represents the duplication of bus routes in transportation community i, L i represents the total length of the road network in traffic zone i, L k represents the length of bus route k, m i represents the number of bus routes in transportation zone i; 3) Road intersection density RID i =N / S i ,i=(1,2,...,l) Among them, RID i represents the road intersection density of traffic zone i, N represents the number of intersections in traffic zone i, S i represents the area of ​​traffic zone i; 4) Road network density RND i =L i / S i ,i=(1,2,...,l) Among them, RND i represents the road network density of traffic zone i, L i represents the total length of the road network in traffic zone i, S i represents the area of ​​traffic zone i; 5) Land use mix Among them, LM i represents the land use mix of traffic zone i, p q is the percentage of the qth type of land area to the corresponding traffic zone area, Q is the total number of land types, s q is the area of ​​the qth type of land.

3. The reliable identification method of residents' travel mode driven by mobile phone signaling data according to claim 1 is characterized in that: The specific calculation steps of travel distance and average travel speed in S3-3 include: 1) Travel duration The representative timestamp of the starting point of a trip is the departure time, the representative timestamp of the destination is the arrival time, and the difference between the two is the trip duration T I ; 2) Travel distance Among them, D I represents the travel distance of user I, n I represents the total number of intersections that user I passes through, d i,i+1 represents the spatial distance between intersection i and intersection i+1; 3) Average travel speed v I =D I / T I Among them, v I represents the average travel speed of user I, T I represents the travel time of user I.

4. The reliable identification method of residents' travel mode driven by mobile phone signaling data according to claim 1 is characterized in that: The specific steps of S5 include: S5-1. Characterization of uncertainty in travel characteristics using Bayesian networks: Statistical analysis is performed on sample data of different travel modes. Based on prior knowledge in the field of transportation, the uncertainty of travel characteristics is expressed in the form of probability, forming a physical characterization scheme for uncertainty, and discretizing continuous characteristic variables into ordered discrete states. Resident travel characteristics include traveler gender, traveler age, travel duration, travel distance, average travel speed, public transportation system service level, road system construction level, and land use level characteristics. S5-2. Divide the mobile phone signaling data with travel mode labels into two data sets, 80% of which is used as a test data set for the node parameters of the Bayesian network model, and 20% is used as a verification data set for the subsequent model effect. Based on the test data set, the node parameters of the Bayesian network model are learned to obtain the conditional probability table of each node in the Bayesian network model, and the travel mode identification model is constructed.

5. The reliable identification method of residents' travel mode driven by mobile phone signaling data according to claim 4 is characterized in that: The formula for learning node parameters of the Bayesian network model is as follows: Among them, θ refers to the maximum posterior probability of θ, D represents the sample data set, and P(D) is the prior distribution of the sample data set.