Airport landside connection traffic mode chain prediction method and system

Abstract

The invention relates to the technical field of traffic behavior prediction, and discloses an airport landside connection traffic mode chain prediction method and system, and the method comprises the steps: obtaining a traffic mode chain selection set based on an airport peripheral traffic network structure and passenger real travel mode selection; social economic attributes, travel feature information and candidate traffic mode scheme attribute data of airport passengers are collected; performing data preprocessing and feature engineering on the collected social and economic attributes, travel feature information and candidate traffic mode scheme attribute data of the airport passengers to obtain effective explanatory variables; and inputting the effective explanatory variables into a pre-constructed XGBoost-based multi-classification prediction model, and outputting a traffic mode chain selection type for predicting airport passengers. According to the invention, complex decision behaviors of airport passengers in multi-mode connection travel can be effectively captured, and a scientific basis is provided for planning and management of an airport ground traffic system.

Description

Technical Field

[0001] This invention belongs to the field of traffic behavior prediction technology, and relates to a method and system for predicting the mode chain of airport landside connection traffic. Background Technology

[0002] With the rapid development of large airport hubs, airport landside transportation systems are characterized by the coexistence of multiple modes, complex connections between modes, and significant heterogeneity in passenger behavior. Traditional research on airport connecting transportation modes often focuses on a single mode of travel (such as taxi, subway, or airport bus) or only on the single segment of the passenger's journey to and from the airport, failing to adequately consider the combined behaviors of multiple modes and chains, and thus making it difficult to reflect the actual entire travel chain of airport access and departure.

[0003] Furthermore, existing models generally employ discrete choice models, which are often limited by model structural assumptions (such as independent and unrelated substitutions, linear additive utility, etc.) when dealing with large-scale, multi-dimensional, and highly correlated explanatory variables, making it difficult to fully capture nonlinear relationships and interaction effects between variables. With the increasing diversity of multimodal travel service providers and passenger demanders, traditional statistical methods are relatively lacking in prediction accuracy, adaptability, and their ability to explain complex behavioral mechanisms. Chinese Patent CN117669837A discloses a method and system for predicting airport landside transfer transportation mode selection. This prediction method uses an RBF neural network model to determine the importance of passenger air travel attribute feature variables, and constructs a Logit prediction model based on the filtered and eliminated passenger air travel attribute feature variables to predict passenger transfer transportation mode selection behavior. While this prediction method uses neural networks for feature importance filtering, the established Logit prediction model focuses on a single travel mode, failing to consider the connection methods of multimodal travel, and struggles to overcome the limitations of discrete choice models in handling large-scale problems, model assumption limitations, and low prediction accuracy.

[0004] In summary, existing inventions and technologies struggle to handle high-dimensional explanatory variables and adapt to complex nonlinear relationships, often focusing on predicting a single connection mode and lacking a predictive method for airport landside connection transportation mode chains with good predictive performance. Furthermore, existing methods lack the ability to explain the internal mechanisms of the model, making it difficult to analyze the contribution of key feature variables to airport passenger connection mode chain selection behavior. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for predicting airport landside connecting transportation modes. This method and system can predict airport landside connecting transportation modes, and no longer focuses on a single connecting transportation mode. Instead, it comprehensively considers the entire process of travel chain modeling for single-mode connecting travel and multi-mode combined connecting travel. It can effectively capture the complex decision-making behavior of airport passengers in multi-mode connecting travel, and provide a scientific basis for the planning and management of airport ground transportation systems.

[0006] To solve the above-mentioned technical problems, the present invention is implemented using the following technical solution.

[0007] In a first aspect, the present invention proposes a method for predicting the mode chain of airport landside access transportation, including:

[0008] Step S1: Construct the transportation mode chain selection set:

[0009] Based on the airport's surrounding transportation network structure and passengers' actual travel mode choices (passengers' actual travel mode choices refer to the transportation modes passengers use throughout their journey from their origin to the airport), we summarize the types of available airport landside transportation mode chains. This yields a set of transportation mode chain choices, including transportation modes such as private cars, taxis / ride-hailing services, subways, airport buses, and high-speed rail, which are formed independently or in combination, such as the high-speed rail-private vehicle intermodal (HSR-PV) combination mode.

[0010] Furthermore, the transportation mode chain includes private car single mode ( Private vehicle unimodal (PV), taxi / ride-hailing single mode ( For-hire Vehicle unimodal (FHV), metro single mode ( Metro unimodal (M), Metro-private car combination mode ( Metro - private vehicle intermodal (M-PV), airport bus-private car combination mode ( Airport coach - private vehicle intermodal (AC-PV), high-speed rail-private car combination mode ( High-speed rail - private vehicle intermodal (HSR-PV), metro-taxi / ride-hailing combination mode ( Metro - For-hire Vehicle intermodal (M-FHV), airport bus - taxi / ride-hailing combination mode ( Airport coach - For-hireVehicle intermodal, AC-FHV), high-speed rail - taxi / ride-hailing combination mode ( Transportation chains such as High-speed rail-For-hire Vehicle intermodal (HSR-FHV).

[0011] S2. Standardized Data Collection:

[0012] S2-a1. Collect data on the socioeconomic attributes of airport passengers (such as gender) through offline questionnaire distribution and online questionnaire distribution via a questionnaire platform. ,age Educational background ,Profession Income level Family car ownership Intercity travel frequency Use of airport frequencies (etc.) and travel characteristic attributes (such as travel purpose) Airfare Flight departure time , number of companions And whether or not you are bringing large luggage. wait)

[0013] S2-a2: Collect attributes of candidate transportation options through multiple network platforms, including the number of subway lines reaching the airport. Number of bus routes that reach the airport Travel time from the departure point to the airport or transfer point for each candidate mode of transportation Costs of each candidate mode of transportation Specifically, the following methods are used: obtaining airport bus routes, stops, and intervals from the airport's official website; obtaining high-speed rail timetables and connection information from online ticketing platforms that sell both air and train tickets; and obtaining private car travel times, subway accessibility, number of bus routes, and fares from the open interface of Gaode Maps.

[0014] S3. Data Preprocessing and Feature Engineering:

[0015] The raw data collected in step S2 are systematically cleaned and filtered to obtain effective explanatory variables, specifically including:

[0016] S3-a1, Missing value completion and outlier removal: Invalid or incomplete questionnaires were removed;

[0017] S3-a2, Eliminating Limited Samples: Considering the negative impact of limited samples on model performance, all candidate transportation modes with fewer than 30 samples were excluded.

[0018] S3-a3. Perform correlation and variance inflation factor (VIF) tests on all explanatory variables to detect correlation and multicollinearity among independent variables. Explanatory variables with VIF > 10 or significant correlation are removed to obtain effective explanatory variables, thereby improving model stability and predictive performance. The explanatory variables refer to socioeconomic attributes, travel characteristic information, and candidate transportation mode scheme attribute data. The effective explanatory variables refer to the preprocessed socioeconomic attributes, travel characteristic information, and candidate transportation mode scheme attribute data.

[0019] S4. Construct a multi-class prediction model based on XGBoost:

[0020] S4-a1: A framework for a multi-class prediction model is established using the XGBoost gradient boosting tree model. Effective explanatory variables, such as the socioeconomic attributes of airport passengers, travel characteristics, and candidate transportation mode attribute data, are input into the multi-class prediction model to predict the choice of connecting transportation mode chains for airport hub passengers. The effective explanatory variables obtained after preprocessing in step S3 are divided into training and test sets in a 70:30 ratio. Grid search is used to optimize hyperparameters, and five-fold cross-validation is used to improve generalization ability. Parameter tuning variables include maximum tree depth (max_depth), number of weak learners (n_estimators), learning rate (learning_rate), subsample ratio (subsample), and feature sampling ratio (colsample_bytree).

[0021] S4-a2: After each calibration of the multi-class prediction model, output the SHAP value of the variable importance, delete variables with low importance, and then test the prediction performance until an ideal balance is achieved between the number of input variables and the model's prediction performance. SHAP interpretation method:

[0022] Based on the Shapley value in cooperative game theory, an average marginal contribution to the prediction results is assigned to each effective explanatory variable. .

[0023] ;

[0024] in, This indicates that the multi-class prediction model does not contain effective explanatory variables. A subset of effective explanatory variables The number of valid explanatory variables used. Indicates the first One valid explanatory variable, This indicates that the set of valid explanatory variables is The output of a multi-class prediction model. The SHAP method can intuitively display the decomposition of variable importance, directional influence, and individual prediction contribution, making up for the weak interpretability of traditional models.

[0025] S5, Traffic Mode Chain Prediction and Performance Evaluation

[0026] The trained model is used to predict the mode chain of new passenger data, and the accuracy is measured. Recall rate F1 score and accuracy The model performance is evaluated using multiple indicators. The airport landside connection transportation mode chain prediction studied in this invention is a multi-class classification problem. The above indicators for each category can be calculated separately, and then the macro average and weighted average of the multi-class model can be obtained by arithmetic average or weighted average, respectively.

[0027] In binary classification problems, True Positive ( ) indicates that a positive sample was successfully predicted as positive, TrueNegative( ) indicates that a negative sample was successfully predicted as negative, False Positive ( ): Mispreting a negative sample as positive, False Negative ( The following are the calculation methods for the evaluation indicators when positive samples are incorrectly predicted as negative:

[0028] S5-a1, Calculate the accuracy for each category ( Accuracy refers to the proportion of samples correctly classified by the model, that is, the ratio of the number of correctly predicted samples to the total number of samples. It is calculated as follows:

[0029] ;

[0030] S5-a2, Calculate the precision for each category ( Precision refers to the proportion of samples that are actually positive out of all samples predicted as positive by the model. It measures the accuracy of the model on samples that are predicted to be positive, and is calculated as follows:

[0031] ;

[0032] S5-a3, Calculate the recall rate for each category ( Recall is the proportion of samples that are actually positive, which are predicted as positive by the model. It measures the model's ability to identify positive samples, and is calculated as follows:

[0033] ;

[0034] S5-a4, Calculate for each category The harmonic mean of precision and recall is a combined measure of both the model's precision and recall. Because the harmonic mean reflects more of the smaller value between the two values ​​in the results, it... The assessment is more accurate for imbalanced category problems, and the calculation method is as follows:

[0035] ;

[0036] S5-a5, Calculate the macro average and the weighted average:

[0037] Macro averaging calculates the index for each choice prediction separately and then averages them to give all choice predictions equal weight; weighted averaging, on the other hand, considers the number of samples for each choice when calculating the index.

[0038] Secondly, this invention proposes an airport landside shuttle mode chain prediction system to implement the aforementioned airport landside shuttle mode chain prediction method, comprising:

[0039] A transportation mode chain selection set construction model is configured to obtain a transportation mode chain selection set based on the transportation network structure around the airport and the actual travel mode choices of passengers.

[0040] The data standardization collection module is configured to collect socioeconomic attributes, travel characteristics information, and candidate transportation mode scheme attribute data of airport passengers.

[0041] The data preprocessing and feature engineering module is configured to perform data preprocessing and feature engineering on the collected socioeconomic attributes, travel characteristics information, and candidate transportation mode scheme attribute data of airport passengers to obtain effective explanatory variables;

[0042] The multi-class prediction model module is configured to input valid explanatory variables into a pre-built XGBoost-based multi-class prediction model and output the predicted mode chain selection type for airport passengers.

[0043] Thirdly, the present invention proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-described airport landside connection transportation mode chain prediction method.

[0044] Fourthly, the present invention provides a computer device comprising:

[0045] Memory, used to store computer programs;

[0046] A processor is used to execute the computer program to implement the steps of the above-described airport landside connection transportation mode chain prediction method.

[0047] Fifthly, the present invention proposes a computer program product, comprising a computer program, characterized in that: when the computer program is executed by a processor, it implements the steps of the above-mentioned airport landside connection transportation mode chain prediction method.

[0048] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0049] (1) This invention realizes the prediction of airport landside connecting transportation mode chain. It no longer focuses on a single connecting transportation mode, but comprehensively considers the whole process of single-mode connecting travel and multi-mode combined connecting travel, which can effectively capture the complex decision-making behavior of airport passengers in multi-mode connecting travel, and provide a scientific basis for the planning and management of airport ground transportation system.

[0050] (2) The XGBoost and other machine learning models of the present invention have overcome the limitations of discrete choice models, such as difficulty in handling complex nonlinear relationships, large model assumptions, and relatively poor prediction performance.

[0051] (3) The present invention introduces the SHAP interpretability method, which enables the machine learning model output to have good interpretability and can identify and filter the key factors that affect the choice of airport passenger connecting transportation mode chain.

[0052] (4) The connection traffic mode chain prediction method involved in this invention has high versatility and scalability. It is not only applicable to the access and outbound traffic mode chain prediction of large hub airports, but can also be extended to other scenarios such as railway integrated transportation hub connection. Attached Figure Description

[0053] Figure 1 This is a schematic diagram of the overall logic flow of the prediction method in Embodiment 1 of the present invention;

[0054] Figure 2 This is a SHAP diagram showing the impact of explanatory variables on the prediction of multi-mode access for metro-private vehicles (M-PV) in Embodiment 1 of the present invention. Detailed Implementation

[0055] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0056] The term "and / or" simply describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0057] Example 1

[0058] This embodiment provides a method for predicting the mode chain of airport landside access transportation, such as... Figure 1 As shown, the process includes constructing a transportation mode chain selection set, standardized data collection, data preprocessing and feature engineering, constructing a multi-class prediction model based on XGBoost, and transportation mode chain prediction and performance evaluation; the specific steps are as follows:

[0059] S1. Construct a transportation mode chain selection set. Based on the structural characteristics of the transportation network around the airport, design a questionnaire. Based on the actual travel mode selection of passengers in the questionnaire, extract and summarize several typical single-mode and multi-mode transportation mode chain types that passengers can choose during the airport landside access and departure processes as the transportation mode chain selection set.

[0060] This embodiment uses an airport as a case study to analyze the choice behavior of airport landside connecting transportation modes. Based on the current situation survey, passengers can reach this airport through various ground transportation modes, including airport buses, subways, high-speed rail, taxis / ride-hailing services, and private cars.

[0061] This embodiment summarizes the optional airport landside access transportation mode chain types. Including but not limited to private car single mode ( Private vehicle unimodal (PV), taxi / ride-hailing single mode ( For-hire Vehicle Unimodal (FHV), Metro Single Mode ( Metro unimodal (M) and combined transportation chains that combine multiple modes of transport, such as the metro-car combination mode ( Metro-private vehicle intermodal (M-PV) and airport bus-private car combination mode ( Airport coach-private vehicle intermodal (AC-PV), high-speed rail-private car combination mode ( High-speed rail - private vehicle intermodal (HSR-PV), subway - taxi / ride-hailing combination mode ( Metro - For-hire Vehicle intermodal (M-FHV), airport bus - taxi / ride-hailing combination mode ( Airport coach - For-hire Vehicle intermodal (AC-FHV), high-speed rail - taxi / ride-hailing combination mode ( Nine types of rail transport chains (High-speed rail - For-hire Vehicle intermodal, HSR-FHV).

[0062] S2, Standardized Data Collection.

[0063] This embodiment collects socioeconomic attributes and travel characteristics information of airport passengers through questionnaire surveys, and obtains candidate transportation scheme attribute data such as travel time, cost and operation service attributes of each candidate transportation scheme through multiple network platforms.

[0064] S2-a1, Offline Questionnaire and Online Survey: In this embodiment, an on-site questionnaire survey was conducted in the domestic departure hall of an airport to obtain complete connecting travel information from passengers. In addition, this embodiment also distributed questionnaires online through a questionnaire platform to collect connecting travel information from domestic passengers who had departed from this airport. The questionnaire included: passenger socioeconomic attributes and travel characteristics data, including gender. ,age Educational background ,Profession Income level Family car ownership Intercity travel frequency Use of airport frequencies Purpose of travel Airfare Flight departure time , number of companions And whether or not you are bringing large luggage. wait.

[0065] S2-a2, data collection of candidate transportation mode scheme attributes from multiple network platforms, including the number of accessible subway lines. Number of accessible bus routes Travel time from the departure point to the airport or transfer point for each candidate mode of transportation Cost of alternative modes of transportation Specifically, the following methods were used: obtaining airport bus routes, stops, and intervals from the airport's official website; obtaining high-speed rail timetables and connection information from online ticketing platforms that sell both air and train tickets; and utilizing online platforms to obtain information on private car travel times, subway accessibility, number of bus routes, and fares. Furthermore, this invention also utilized online platforms to extract the number of bus routes within an 800-meter radius of the airport. and number of subway lines .

[0066] S3. Data preprocessing and feature engineering, including missing value handling, outlier removal, variable correlation analysis, and variance inflation factor test, to screen for irrelevant and effective explanatory variables.

[0067] S3-a1, Missing value completion and outlier removal: To ensure the validity of the data, we screened the questionnaires and removed invalid or incomplete questionnaires.

[0068] S3-a2, Eliminating Limited Samples: Considering the negative impact of limited samples on the performance of multi-class prediction models, all candidate transportation modes with fewer than 30 samples were excluded.

[0069] S3-a3, variance inflation factor (VIF) analysis, removes explanatory variables with VIF>10 or significant correlation, improving model stability and predictive performance.

[0070] After data preprocessing and feature engineering, 844 valid samples were retained. Descriptive statistics for some feature variables are as follows:

[0071] gender : Male (47.9%), Female (52.1%), Age Ages: 18-29 (58.9%), 30-39 (29.4%), 40-49 (9.6%), 50-59 (1.9%), 60 and above (0.2%); Educational background. High school or below (7.3%); Bachelor's degree (61.7%); Master's degree or above (30.9%); Occupation Students (38.0%), government employees (21.8%), corporate employees (24.4%), self-employed / business owners (8.6%), other occupations (7.1%), income level ≤3000 yuan (28.0%); 3000-6000 yuan (13.9%); 6001-10000 yuan (18.8%); 10001-20000 yuan (26.3%); 20001-50000 yuan (9.2%); ≥50001 yuan (3.8%); family car ownership 0 vehicles (15.3%); 1 vehicle (50.5%); ≥2 vehicles (34.2%); Intercity travel frequency Airport frequency usage: 0 times / year (1.3%); 1-2 times / year (25.8%); 3-4 times / year (30.5%); 5-6 times / year (15.3%); 7-8 times / year (5.6%); 9-10 times / year (4.5%); ≥10 times / year (17.1%) : 0 times (17.7%); 1-2 times (36.7%); 3-4 times (25.1%); 5-6 times (8.6%); ≥7 times (11.8%), Purpose of travel Business travel (40.6%), non-business travel (59.4%), airfare ≤500 yuan (8.2%); 501-1000 yuan (41.2%); 1000-2000 yuan (38.2%); 2000-3000 yuan (8.1%); 3000-4000 yuan (2.4%); ≥4000 yuan (2.0%); Flight departure time Non-early morning / late evening flights (10:00 AM - 9:00 PM) (79.6%); Early morning / late evening flights (before 10:00 AM / after 9:00 PM) (20.4%), number of passengers. : 0 people (41.1%); 1 person (20.3%); 2 people (30.6%); 3 people (7.2%); ≥4 people (0.8%) and whether they brought large luggage. : 0 pieces or carry-on baggage only (33.9%); 1 piece (65.3%); ≥2 pieces (0.8%).

[0072] S4. Construct a multi-class prediction model based on XGBoost to obtain the multi-class prediction model. Input the preprocessed variables into the multi-class prediction model to predict the passenger's mode of transport chain selection type. Use grid search and cross-validation methods to determine the optimal hyperparameter combination of the multi-class prediction model, and identify and screen important variables based on SHAP graph.

[0073] S4-a1: A multi-class prediction framework for machine learning is established using the XGBoost gradient boosting tree model, resulting in a multi-class prediction model based on XGBoost. Passenger features are input into the multi-class prediction model to predict the choice of passenger connecting transportation modes at the airport hub. The multi-class prediction model is trained using a 70% training set and a 30% test set. Grid search is used to optimize hyperparameters, and five-fold cross-validation is used to improve generalization ability. Parameter tuning variables include maximum tree depth (max_depth = 5), number of weak learners (n_estimators = 50), learning rate (learning_rate = 0.2), subsample ratio (subsample = 0.6), and feature sampling ratio (colsample_bytree = 0.6).

[0074] S4-a2: After each calibration of the multi-class prediction model, output the SHAP value of the variable importance, delete variables with low importance, and then test the prediction performance until an ideal balance is achieved between the number of input variables and the model's prediction performance. For example... Figure 2 As shown, the impact of explanatory variables on the prediction of multimodal access for metro-private vehicles (M-PV) is illustrated.

[0075] in, Figure 2 The data includes: "Travel time from the departure point to the nearest high-speed rail station by private car", "Travel time from the departure point to the nearest subway station by private car", "Travel time from the departure point to the nearest airport bus station by private car", "Travel time from the nearest subway station to the airport by subway and airport bus", "Number of bus routes within 800 meters of the departure point", "Travel time from the departure point to the airport by private car", "Number of subway routes within 800 meters of the departure point", "Travel time from the nearest airport bus stop to the airport", and "Fare of airport bus from the nearest stop to the airport". These are all attribute data for candidate transportation modes.

[0076] S5. Mode Chain Prediction and Performance Evaluation: A trained multi-class prediction model is used to predict the mode chain of new passenger samples. Precision, recall, and other metrics are then evaluated. The performance of multi-class prediction models is evaluated using various classification metrics such as accuracy.

[0077] Using a trained multi-class prediction model, traffic mode chain prediction is performed on airport passenger data, and the accuracy is measured. Recall rate F1 score and accuracy The performance of the multi-class prediction model is evaluated using various classification indicators. In this embodiment, the prediction of airport landside connection transportation mode chains is an imbalanced multi-class problem. The aforementioned indicators for each category can be calculated separately, and then the macro average and weighted average of the multi-class prediction model can be obtained respectively by arithmetic average or weighted average.

[0078] In binary classification problems, True Positive ( ) indicates that a positive sample was successfully predicted as positive; TrueNegative ( ) indicates that a negative sample was successfully predicted as negative; False Positive ( ) indicates that a negative sample was incorrectly predicted as positive; False Negative ( This indicates that a positive sample was incorrectly predicted as negative. The calculation methods for each evaluation index are as follows:

[0079] S5-a1, Calculate the accuracy for each category ( Accuracy refers to the proportion of samples correctly classified by a multi-class prediction model in all samples, that is, the ratio of the number of correctly predicted samples to the total number of samples.

[0080] S5-a2, Calculate the precision for each category ( Precision is the proportion of samples that a multi-class classification model predicts as positive, and which are actually positive. It measures the accuracy of the model on samples that are predicted to be positive.

[0081] S5-a3, Calculate the recall rate for each category ( Recall is the proportion of samples that are actually positive, which are predicted as positive by a multi-class classification model. It measures the model's ability to identify positive samples.

[0082] S5-a4, Calculate for each category The harmonic mean of precision and recall is a combined measure of both the precision and recall of a multi-class prediction model. Because the harmonic mean reflects more of the smaller of the two values ​​in the result, it... It is more accurate for assessing imbalanced category problems.

[0083] S5-a5, Calculate the macro average and the weighted average:

[0084] Macro averaging calculates the index for each choice prediction separately and then averages them to give all choice predictions equal weight; weighted averaging, on the other hand, considers the number of samples for each choice when calculating the index.

[0085] As shown in Table 1, after calculation, the accuracy of this embodiment in the 254 test set samples is 0.88, the macro-average precision is 0.92, the macro-average recall is 0.83, and the macro-average... It is 0.87; weighted average precision, weighted average recall, weighted average Both are 0.88.

[0086] Table 1. Model Performance on the Test Set

[0087]

[0088] Example 2

[0089] Based on the same inventive concept as Embodiment 1, this embodiment introduces an airport landside connection transportation mode chain prediction system, including:

[0090] A transportation mode chain selection set construction model is configured to obtain a transportation mode chain selection set based on the transportation network structure around the airport and the actual travel mode choices of passengers.

[0091] The data standardization collection module is configured to collect socioeconomic attributes, travel characteristics information, and candidate transportation mode scheme attribute data of airport passengers.

[0092] The data preprocessing and feature engineering module is configured to perform data preprocessing and feature engineering on the collected socioeconomic attributes, travel characteristics information, and candidate transportation mode scheme attribute data of airport passengers to obtain effective explanatory variables;

[0093] The multi-class prediction model module is configured to input valid explanatory variables into a pre-built XGBoost-based multi-class prediction model and output the predicted mode chain selection type for airport passengers.

[0094] Example 3

[0095] Based on the same inventive concept as other embodiments, this embodiment introduces a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described airport landside connection transportation mode chain prediction method.

[0096] Example 4

[0097] Based on the same inventive concept as other embodiments, this embodiment introduces a computer device, including: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the above-described airport landside connection transportation mode chain prediction method.

[0098] Example 5

[0099] Based on the same inventive concept as other embodiments, this embodiment introduces a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described airport landside connection transportation mode chain prediction method.

[0100] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0101] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0102] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0103] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0104] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other modifications under the guidance of the present invention, and these modifications are all within the protection scope of the present invention.

Claims

1. An airport landside access mode chain prediction method, characterized by, The method comprises the following steps: Based on the structure of the airport surrounding traffic network and the real travel mode selection of passengers, a traffic mode chain selection set is obtained; Collecting the social and economic attributes, travel characteristics information and candidate traffic mode scheme attribute data of airport passengers; Data preprocessing and feature engineering are performed on the collected social and economic attributes, travel characteristics information and candidate traffic mode scheme attribute data of airport passengers to obtain effective explanatory variables; The effective explanatory variables are input into a pre-constructed XGBoost-based multi-classification prediction model to output the predicted traffic mode chain selection type of airport passengers.

2. The method of claim 1, wherein: The traffic mode chain is a traffic mode chain formed by private cars, taxis, subways, airport buses and high-speed rails independently or in combination.

3. The method of claim 1, wherein: The collected social and economic attributes, travel characteristics information and candidate traffic mode scheme attribute data of airport passengers include: Information on the socioeconomic attributes and travel characteristics of airport passengers was collected through offline questionnaires and online surveys; the socioeconomic attributes included gender. ,age Educational background ,Profession Income level Family car ownership Intercity travel frequency Use of airport frequencies The travel characteristic information includes the travel purpose. Airfare Flight departure time , number of companions And whether to bring large luggage ; Collecting candidate traffic mode scheme attribute data through a network platform, the candidate traffic mode scheme attribute data including a number of subway lines capable of reaching the airport , a number of bus lines capable of reaching the airport , travel times of the candidate traffic modes from a departure location to the airport or to a transfer node , and fees of the candidate traffic modes .

4. The method of claim 1, wherein: The data preprocessing and feature engineering performed on the collected social and economic attributes, travel characteristics information and candidate traffic mode scheme attribute data of airport passengers include: Missing value processing, outlier removal, variable correlation analysis and variance inflation factor test are performed to screen out irrelevant effective explanatory variables; the effective explanatory variables refer to the pre-processed social and economic attributes, travel characteristics information and candidate traffic mode scheme attribute data.

5. The airport land-side transfer traffic mode chain prediction method according to claim 1, wherein: The construction method of the XGBoost-based multi-classification prediction model comprises the following steps: A gradient boosting tree model XGBoost is used to establish a multi-classification prediction model framework; A grid search and cross-validation method is used to determine the optimal hyperparameters of the multi-classification prediction model to realize the calibration of the multi-classification prediction model.

6. The method of claim 5, wherein: The construction method of the XGBoost-based multi-classification prediction model further comprises the steps of identifying and screening important variables based on SHAP graphs, specifically: After each calibration of the multi-classification prediction model, the SHAP value variable importance is output, the variables with low importance are deleted, and the prediction performance is tested until the ideal balance of the number of input variables and the model prediction performance is achieved; the SHAP explanation method is as follows: assigning, based on Shapley values in cooperative game theory, an average marginal contribution to a prediction result for each valid explanatory variable ; ； wherein, represents a subset of valid explanatory variables not contained in the multiclass prediction model , is the number of valid explanatory variables used, represents the th valid explanatory variable, represents the output of the model when the set of valid explanatory variables is .

7. An airport landside access mode chain prediction system for implementing the airport landside access mode chain prediction method according to any one of claims 1 to 6, characterized by The method comprises the following steps: A traffic mode chain selection set construction model is configured to obtain a traffic mode chain selection set based on the structure of the airport surrounding traffic network and the real travel mode selection of passengers; A data standardization collection module is configured to collect the social and economic attributes, travel characteristics information and candidate traffic mode scheme attribute data of airport passengers; A data preprocessing and feature engineering module is configured to perform data preprocessing and feature engineering on the collected social and economic attributes, travel characteristics information and candidate traffic mode scheme attribute data of airport passengers to obtain effective explanatory variables; A multi-classification prediction model module is configured to input the effective explanatory variables into a pre-constructed XGBoost-based multi-classification prediction model to output the predicted traffic mode chain selection type of airport passengers.

8. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by a processor to realize the steps of the airport land-side transfer traffic mode chain prediction method of any one of claims 1-6.

9. A computer device, comprising: The method comprises the following steps: A memory for storing a computer program; A processor for executing the computer program to implement the steps of the method for predicting the airport landside access mode chain according to any one of claims 1-6.

10. A computer program product comprising a computer program, characterized in that: The computer program is executed by the processor to implement the steps of the method for predicting the airport landside access mode chain according to any one of claims 1-6.

Images