CN-122022047-A - Multi-dimensional feature-based connection passenger flow prediction method and device

CN122022047ACN 122022047 ACN122022047 ACN 122022047ACN-122022047-A

Abstract

The invention relates to the technical field of data analysis and discloses a method and a device for predicting connected passenger flow based on multidimensional features, wherein the method comprises the steps of obtaining user riding data corresponding to a public transportation network and passenger flow influence data associated with the user riding data; the method comprises the steps of preprocessing user riding data to obtain target riding data, calculating historical riding data based on a riding passenger flow prediction formula and according to the target riding data, extracting characteristics of passenger flow influence data and historical riding passenger flow data based on a multi-dimensional characteristic processing model to obtain multi-dimensional characteristics related to the historical riding passenger flow data, and calculating predicted riding passenger flow of any riding station according to the multi-dimensional characteristics based on a dynamic generalized additive model. Therefore, the implementation of the method can realize the dynamic prediction of the bus subway connection passenger flow on the basis of improving the analysis comprehensiveness of the multidimensional features of the bus subway connection, thereby improving the prediction accuracy of the bus subway connection passenger flow.

Inventors

YU HONGLING
ZENG YE
Luo Sishu
ZHOU LONGTAO

Assignees

广州羊城通有限公司

Dates

Publication Date: 20260512
Application Date: 20260202

Claims (10)

1. A method for predicting a connected passenger flow based on multidimensional features, the method comprising: Acquiring user riding data corresponding to a public transportation network and passenger flow influence data associated with the user riding data, wherein the user riding data comprises riding data of a plurality of users; Preprocessing the user riding data to obtain target riding data, wherein the preprocessing comprises time alignment processing, outlier processing and transfer behavior screening processing; Based on a connection passenger flow prediction formula, according to the target riding data, historical connection passenger flow data are counted; Performing feature extraction operation on the passenger flow influence data and the historical connection passenger flow data based on a multi-dimensional feature processing model to obtain multi-dimensional features associated with the historical connection passenger flow data, wherein the multi-dimensional features comprise time features, space features and historical passenger flow features; Based on a dynamic generalized additive model, calculating predicted connection passenger flow of any connection station according to the multi-dimensional characteristics, wherein the dynamic generalized additive model is used for fusing the multi-dimensional characteristics, and each connection station consists of a connection subway station and a connection bus station which have connection relations.
2. The method for predicting the docking passenger flow based on the multidimensional feature according to claim 1, wherein the riding data of each user comprises an account identifier of the user, subway riding data of the user and bus riding data of the user, wherein the subway riding data comprises the arrival time of the user at a plurality of first subway stations and the departure time of the user at a plurality of second subway stations; The preprocessing the user riding data to obtain target riding data comprises the following steps: Performing time alignment processing on the user riding data to obtain time alignment riding data; Performing outlier processing on the time-aligned riding data to obtain cleaning and completing riding data; And screening riding data meeting preset effective transfer behavior conditions from the cleaning and completing riding data to serve as target riding data.
3. The method for predicting the connection passenger flow based on the multidimensional feature according to claim 2, wherein the calculating historical connection passenger flow data based on the connection passenger flow prediction formula according to the target riding data comprises the following steps: Counting the connected passenger flow data about the connected station in the target riding data based on a connected passenger flow prediction formula and according to a transfer time window corresponding to the effective transfer behavior condition and all account identifications, wherein the connected passenger flow data comprises first connected passenger flow data and second connected passenger flow data, the first connected passenger flow data is the number of passengers transferred to the connected subway station by the connected bus station in the time window length corresponding to the transfer time window, and the second connected passenger flow data is the number of passengers transferred to the connected subway station by the connected subway station in the time window length corresponding to the transfer time window; Generating a connection passenger flow curve corresponding to each connection station based on preset time sequence arrangement conditions according to the connection passenger flow data, wherein the connection passenger flow curve records connection passenger flows corresponding to a plurality of moments; the historical connection passenger flow data comprise connection passenger flow data corresponding to each connection station and connection passenger flow curves corresponding to each connection station.
4. A method of predicting connected passenger flow based on multi-dimensional features according to any one of claims 1-3, wherein the performing feature extraction operations on the passenger flow influence data and the historical connected passenger flow data based on a multi-dimensional feature processing model to obtain multi-dimensional features associated with the historical connected passenger flow data comprises: Performing time feature processing operation on the time data based on a time dimension feature model to obtain time features, wherein the time features comprise week features, hour features, minute features and holiday features; Based on a space dimension feature model, performing space feature processing operation on the space data to obtain space features, wherein the space features comprise weather features, emergency features and peripheral activity features; and based on the historical passenger flow characteristic model, executing passenger flow characteristic processing operation on the historical connection passenger flow data to obtain historical passenger flow characteristics.
5. The multi-dimensional feature based docking passenger flow prediction method of claim 4, wherein the time data comprises week data, hour data, minute data, and holiday data; The time dimension feature model is based on, and the time feature processing operation is executed on the time data to obtain time features, and the method comprises the following steps: Calculating a week feature from the week data based on a first periodic B-spline function, wherein the first periodic B-spline function is a cubic B-spline basis function; Calculating an hour feature from the hour data based on a second periodic B-spline function, wherein the second periodic B-spline function is another cubic B-spline basis function different from the first periodic B-spline function; calculating a minute characteristic according to the minute data based on a minute characteristic mapping formula; and analyzing the holiday data based on a holiday factor function to obtain holiday characteristics, wherein the holiday factor function is provided with an indicating variable related to holidays.
6. The multi-dimensional feature based docking passenger flow prediction method of claim 4, wherein the spatial data comprises weather data, emergency data, and peripheral activity data; The spatial dimension feature model is based on, and spatial feature processing operation is performed on the spatial data to obtain spatial features, including: Classifying the weather data based on preset weather classification conditions to obtain weather classification results corresponding to the weather data; Analyzing the weather classification result based on a weather factor function to obtain weather characteristics, wherein the weather factor function is provided with a coefficient to be estimated, and the coefficient to be estimated is used for representing the influence degree of the weather classification result on the connected passenger flow; analyzing the emergency data based on an emergency factor function to obtain an emergency characteristic, wherein the emergency factor function is provided with an indication variable of an emergency corresponding to any one of the docking stations; and analyzing the peripheral activity data based on a peripheral activity factor function to obtain peripheral activity characteristics, wherein the peripheral activity factor function is provided with an indication variable related to the peripheral activity corresponding to any docking station.
7. The method for predicting connected passenger flow based on multidimensional features according to claim 4, wherein the performing a passenger flow feature processing operation on the historical connected passenger flow data based on the historical passenger flow feature model to obtain historical passenger flow features comprises: The method comprises the steps of calculating historical passenger flow characteristics corresponding to historical connection passenger flow data based on a pre-constructed quadratic polynomial regression formula, wherein the quadratic polynomial regression formula comprises a linear term, a nonlinear term and an intersection term, the linear term is used for capturing characteristic linear relations corresponding to the historical connection passenger flow data, the nonlinear term is used for capturing characteristic nonlinear relations corresponding to the historical connection passenger flow data, and the intersection term is used for capturing interaction between characteristics corresponding to adjacent passenger flow data in the historical connection passenger flow data.
8. A docking passenger flow prediction device based on multidimensional features, the device comprising: The system comprises an acquisition module, a public transportation network, a storage module and a storage module, wherein the acquisition module is used for acquiring user riding data corresponding to the public transportation network and passenger flow influence data associated with the user riding data, wherein the user riding data comprises riding data of a plurality of users; The preprocessing module is used for preprocessing the user riding data to obtain target riding data, wherein the preprocessing comprises time alignment processing, outlier processing and transfer behavior screening processing; the statistics module is used for counting historical connection passenger flow data according to the target riding data based on a connection passenger flow prediction formula; The feature extraction module is used for executing feature extraction operation on the passenger flow influence data and the historical connection passenger flow data based on a multi-dimensional feature processing model to obtain multi-dimensional features associated with the historical connection passenger flow data, wherein the multi-dimensional features comprise time features, space features and historical passenger flow features; The feature fusion module is used for calculating the predicted connection passenger flow of any connection station according to the multi-dimensional features based on the dynamic generalized additive model, wherein the dynamic generalized additive model is used for fusing the multi-dimensional features, and each connection station consists of a connection subway station and a connection bus station which have connection relations.
9. A docking passenger flow prediction device based on multidimensional features, the device comprising: a memory storing executable program code; a processor coupled to the memory; The processor invokes the executable program code stored in the memory to perform the docking passenger flow prediction method based on the multi-dimensional features of any one of claims 1-7.
10. A computer storage medium storing computer instructions which, when invoked, are operable to perform the method of docking passenger flow prediction based on multi-dimensional features of any one of claims 1-7.

Description

Multi-dimensional feature-based connection passenger flow prediction method and device Technical Field The invention relates to the technical field of data analysis, in particular to a method and a device for predicting connection passenger flow based on multidimensional features. Background The bus subway connection refers to the transfer behavior of passengers between two public transportation modes of buses and subways, and is a key link of a city multi-mode public transportation system. The efficient and accurate prediction of the connected passenger flow is an important data base for optimizing the layout of the connected stations and coordinating the line transportation capacity, and is important for improving the operation efficiency of a public transportation network and the passenger transfer experience. Existing passenger flow prediction methods rely mainly on historical passenger flow time series analysis or statistical modeling based on generalized linear models (GLM, generalized Linear Model). However, in practice, it is found that the conventional methods often preset the relationship between the influencing factors and the passenger flows to be linear, and meanwhile, the characteristic dimension of the existing passenger flow prediction model is single, focusing on a macroscopic time sequence rule, and dynamic analysis of microscopic passenger flow influencing factors which are easy to mutate cannot be performed, so that the prediction accuracy of the connected passenger flows is low. Therefore, it is important to provide a technical scheme that can realize dynamic prediction of the bus subway connection passenger flow on the basis of improving the analysis comprehensiveness of the multidimensional features of the bus subway connection, so as to improve the prediction accuracy of the bus subway connection passenger flow and facilitate providing more accurate data basis for the operation scheduling optimization of a public transportation network. Disclosure of Invention The invention provides a method and a device for predicting the connected passenger flow based on multidimensional features, which can realize dynamic prediction of the connected passenger flow of the subway on the basis of improving the analysis comprehensiveness of the multidimensional features of the connection of the subway, thereby improving the prediction accuracy of the connected passenger flow of the subway and being beneficial to providing more accurate data basis for the operation scheduling optimization of a public transportation network. In order to solve the technical problems, the first aspect of the invention discloses a method for predicting the connection passenger flow based on multidimensional features, which comprises the following steps: Acquiring user riding data corresponding to a public transportation network and passenger flow influence data associated with the user riding data, wherein the user riding data comprises riding data of a plurality of users; Preprocessing the user riding data to obtain target riding data, wherein the preprocessing comprises time alignment processing, outlier processing and transfer behavior screening processing; Based on a connection passenger flow prediction formula, according to the target riding data, historical connection passenger flow data are counted; Performing feature extraction operation on the passenger flow influence data and the historical connection passenger flow data based on a multi-dimensional feature processing model to obtain multi-dimensional features associated with the historical connection passenger flow data, wherein the multi-dimensional features comprise time features, space features and historical passenger flow features; Based on a dynamic generalized additive model, calculating the predicted connection passenger flow of any connection station according to the multi-dimensional characteristics, wherein the dynamic generalized additive model is used for fusing the multi-dimensional characteristics, and each connection station consists of a connection subway station and a connection bus station which have connection relations. In a first aspect of the present invention, each of the user's boarding data includes an account identifier of the user, subway boarding data of the user, and bus boarding data of the user, wherein the subway boarding data includes an inbound time of the user at a plurality of first subway stations and an outbound time of the user at a plurality of second subway stations; The preprocessing the user riding data to obtain target riding data comprises the following steps: Performing time alignment processing on the user riding data to obtain time alignment riding data; Performing outlier processing on the time-aligned riding data to obtain cleaning and completing riding data; And screening riding data meeting preset effective transfer behavior conditions from the cleaning and completing riding data to serve as target riding data