CN103366224B - Passenger demand prediction system and method based on public transport network - Google Patents
Passenger demand prediction system and method based on public transport network Download PDFInfo
- Publication number
- CN103366224B CN103366224B CN201310293407.3A CN201310293407A CN103366224B CN 103366224 B CN103366224 B CN 103366224B CN 201310293407 A CN201310293407 A CN 201310293407A CN 103366224 B CN103366224 B CN 103366224B
- Authority
- CN
- China
- Prior art keywords
- time
- bus
- model
- passenger demand
- prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a passenger demand prediction system and method based on a public transport network, which comprehensively consider factors such as heterogeneity, burstiness and periodicity, and finally obtain passenger demand prediction in the public transport network through a Poisson model changing along with time, a weighted time-varying Poisson model, a comprehensive autoregressive moving average model and other prediction models and an integration frame based on a sliding window. The predicted passenger demand can provide a more convenient and comfortable bus travel environment for the passengers, such as reducing waiting time of the passengers and avoiding the situation that the buses are excessively crowded or loosened.
Description
Technical Field
The invention relates to a passenger demand prediction system and method based on a public transport network.
Background
With the rapid development of economy, the transportation industry has been developed rapidly, but the traffic conditions are continuously worsened, and a series of traffic problems are generated: the phenomena of traffic congestion and road blockage become more serious, traffic accidents frequently occur, and the energy consumption and the environmental pollution brought by the traffic accidents also attract more and more general social attention. Public transportation, especially bus service, can effectively alleviate these problems. The public transportation service can effectively utilize the existing traffic facilities, reduce traffic load and environmental pollution, ensure traffic safety, improve transportation efficiency and improve the convenience and comfort of road users. In addition, buses are widely distributed and cheap, so that the buses are increasingly favored by people in various countries. In 2011, in 500 thousands of residents in singapore, 330 more than ten thousands of residents take buses every day. However, the service quality of the current buses still needs to be improved, and passengers want to shorten waiting time as much as possible when taking the buses and take uncongested buses as much as possible. In fact, overcrowded buses may scare away many passengers, causing them to give up taking the bus. A reasonably balanced bus service should therefore be able to maximize the benefits of both the public transportation company and the passengers. If this equalization is lost, one of two situations can occur: 1) excessive empty vehicles and little passenger demand; 2) passengers with long waiting times and overcrowded buses. Therefore, accurate and real-time passenger demand prediction can help a bus company to determine a reasonable bus departure time interval, and the waiting time of passengers can be reduced, which is urgently needed by people.
However, due to the presence of many uncertainties, the present invention needs to face three challenges: 1) non-uniformity. The requirements of passengers on the public transportation service are different among different stops, different working days and different time periods of the same day; 2) is bursty. The demand of passengers at each bus stop is different, and the demand of passengers at a plurality of bus stops is sudden and can be influenced by a plurality of accidents, such as traffic jam, weather change and the like; 3) and (4) periodicity. Passenger demand for bus service is highly relevant in the same working day of different weeks, and in the morning and evening of the same day. To solve the challenge problems well, the invention provides a passenger demand prediction system and method based on a public transportation network. Based on historical GPS data and bus service data (e.g., passengers' boarding/disembarking stops), a time series histogram of P minutes is established for the passenger demand at each bus stop. The invention adopts famous time series prediction technology, such as a Poisson model changing along with time, a Poisson model changing in weighted time, a comprehensive autoregressive moving average model and the like to predict the passenger demand in the public transportation network.
Efficient passenger demand forecasting will become an important new feature for providing advanced Services in public transportation networks, and is also very useful for other Location Based Services (LBS) applications. However, to date, passenger demand prediction in public transportation networks has never been considered. Vuchic (v.r. Vuchic.: Transit Operating manual. Department of transportation. Pennsylvania, usa. 1976) proposes a method of determining the departure time interval of a bus, thereby providing sufficient transportation capacity. Daganzo (C.F. Daganzo.: A Headway-based Approach to interference Bus Bunching: Systematic Analysis and compatibility. Transp. Part B43: 913-921.2009) proposes an adaptive control mode to Eliminate Bus Bunching and dynamically determine the departure time of a Bus according to real-time Headway information. Zhao et al (j. Zhao, m. Dessouky, s. bukkapatnam.: Optimal Slack Time for Schedule-based transactions operation. trans. sci. 40 (4): 529-539.2006) studied the problem of Optimal Slack Time to minimize the waiting Time of passengers by bus transport vehicle control. Chen (H. Chen.: storage Optimization In calculating Multiple Headways for a Single Bus line of the 35th Annual correlation Symposium (ANS-35). IEEE Computer Society, California, USA 2002) considers the problem of Multiple distances on the same Bus line, and the Optimization model is to maximize the profit of the public transportation company. Yan et al (S.Y. Yan, C.J. Chi, C.H. Tang.: Inter-city Bus Routing and time table Setting under stored demand management. Transp. Res. Part A40: 572-586.2006) studied the Setting model of route and schedule for the random demand of intercity Bus line, however this model did not consider the city Bus line and did not analyze the demand difference in different time periods. None of the above studies have considered the real-time passenger demand for public transportation services, which is the focus of the present invention.
There are currently many relevant studies on the prediction of bus arrival time. Van Hinsbergen et al (c.v. Hinsbergen, j.v. link, h.j. zuylen.: Bayesian Committee of Neural Networks to preliminary Times with configuration intervals. Transportation Research Part C, vol.17, pp. 498-509.2009) fuse Neural Networks into a latency prediction model and solve the problem of selecting an optimal network using bayes theory. Chang et al (H. Chang, D. Park, S. Lee, H. Lee, S. Baek.: Dynamic Multi-interval Bus transition Time Prediction using Bus Transit data. Transportmetric Vol. 6, pp. 19-38.2010) propose a Dynamic model based on nearest neighbor nonparametric regression to predict the Travel Time of multiple paths from the starting station to the ending station. Yu et al (B, Yu, W.Lam, M.L.tam.: Bus Arrival Time Prediction at Bus Stop with Multiple routes. transfer Research Part C Vol. 19, pp. 1157-1170.2011) propose various methods including Support Vector Machine (SVM), Artificial Neural Network (ANN), K-nearest neighbor algorithm (KNN) and Linear Regression (LR) to predict the waiting Time, the result proves that the SVM model is the best for predicting the waiting Time of a Bus Stop with Multiple routes. There is also currently relevant research work on traffic congestion. (A. Lakas and M. Chaqfeh.: A Novel Method for Reducing Road Traffic Congestion communication. In Proceedings of the 6th International Wireless Communications and Mobile Computing Conference, IWCMC, ACM, pp. 16-20.2010) proposed a vehicle-mounted communication system for detecting and warning Traffic Congestion by mining and propagating Road information. The system consists of two parts: one is based on the protocol of flooding to transmit traffic information, and the other is based on the algorithm of dijkstra to dynamically calculate the minimum congestion route of the vehicle. A virtual Traffic light protocol is designed In (M, Ferreira, R, Fernandes, H, Conceico, W, Viriyasitavat, and O, Tonguz, Self-organized Traffic control, In Proceedings of the Annual International Conference on Mobile Computing and Networking, MOBICM, ACM, pp. 85-90.2010), and the Traffic flow of a road intersection is dynamically optimized without any roadside infrastructure. The above method assumes that each vehicle in the road is voluntarily involved in traffic management and is actively providing relevant information. However, many drivers are willing to enjoy the convenience of traffic information and are reluctant to share any information. Because of this selfishness, none of these traffic management systems are feasible. The passenger demand prediction method can help the public transport company to determine a reasonable bus departure time interval, and reduce the waiting time of passengers, thereby reducing or even avoiding traffic jam.
In summary, the main disadvantages of the existing methods are: 1. the current research work does not consider the real-time demand of passengers on the bus service, which is a main factor in the traffic management; 2. much of the current research work is premised on the assumption that all drivers are actively engaged, which is not feasible in most practical applications; 3. the existing research work does not consider the non-uniformity, the burstiness and the periodicity of the passenger demands, which must be considered in the passenger demand analysis, otherwise, the final predicted passenger demand result is influenced; 4. the data of the existing taxi-based research work is incomplete.
Disclosure of Invention
Aiming at the problems in the prior art, the invention aims to provide a passenger demand prediction system and method based on a public transport network. The method comprehensively considers factors such as heterogeneity, burstiness and periodicity, and provides an accurate and real-time passenger demand prediction system and method in the public transport network for public transport companies and passengers.
In order to achieve the purpose, the passenger demand prediction system and method based on the public transport network comprises the following specific steps:
1) starting from the application of an actual public transport network, providing a summary description of a passenger demand prediction method;
2) the factors such as heterogeneity, burstiness and periodicity are comprehensively considered, and three different passenger demand prediction models are respectively provided;
3) a sliding window based framework is proposed to integrate the three prediction models.
Further, the actual public transportation network in step 1) is represented as follows: suppose a bus route contains N (N)
2) Bus stop S = { S1,S2,…,SN}. The first site is a starting site, and the last site is a terminating site. The transit of the bus between stops follows a specific route and a specific timetable. Db ={d1,d2,…,djRepresents the set of destinations for j passengers on b buses getting on at station s. The running time is divided into 4 time periods (5 am to 9 am, 9 am to 1 am, 1 am to 5 pm, and 5 pm to 9 pm) according to the shift change time of the bus. The invention aims to solve the problem of predicting the number of passengers who want to take a b-way bus at a bus stop s at the time t.
Further, the prediction model for the passenger requirement in the step 2) comprises a time-varying poisson model, a weighted time-varying poisson model and a comprehensive autoregressive moving average model.
Further, the time-varying poisson model comprises the steps of: the probability P (n) that a certain bus stop has n buses to stop in a given time meets Poisson distribution and is defined as
In the formula
Representing the average passenger demand for bus service over a fixed period of time, the invention
Is not constant but varies over time. Therefore we consider it as a function of time
Thereby transforming the poisson distribution to non-homogeneous.
Is defined as
Where d (t) represents weekday {1= sunday, 2= monday, … }, and h (t) is the time period to which time t belongs (for example, if every 30 minutes is a time period, time 00:31 is included in the second time period). In addition, the following two equations need to be satisfied
Wherein D is the number of time periods in a day;
is the average ratio of the one-week poisson processes;
indicates relative change on day i (e.g., saturday rates are lower than tuesday);
indicates the relative change in the ith time period on day j (e.g., peak hours);
is a discrete function representing the time-varying passenger demand at bus stop s.
Further, the weighted time-varying poisson model comprises the steps of: the time-varying poisson model only predicts the time-dependent average passenger demand, however the passenger demand is different for each bus stop. In fact, the passenger demands of many public transportation stations are sudden and affected by many accidents, such as traffic jam and weather change. The weighted time-varying poisson model can solve the sudden problem well. The objective is to increase the correlation of the passenger demand in the last week with the passenger demand in the previous weeks. The weight w of the correlation is calculated by the famous time series method-exponential smoothing method-which is defined as
In the formula
Is an average of the passenger demand over the past time period,
is a smoothing factor whose value is used byUser-defined, having a size range of 0<
<1。
Further, the comprehensive autoregressive moving average model comprises the following steps: both previous models assume that there is periodic regularity in the passenger's demand for public transportation services, and in fact the passenger's demand is different between different stops, different work days, and different time periods on the same day. The comprehensive autoregressive moving average model can well simulate and predict univariate time series data such as traffic flow data, short-term prediction problems and the like. This has the advantage of being able to accurately represent different types of time series, such as auto-regressive time series, moving average time series, and combinations thereof. In the synthetic autoregressive moving average model, the predicted values of variables can be viewed as linear functions of historical observations and random errors. In the invention, the time sequence is regarded as the passenger demand quantity changing along with the time on a specific bus stop s, so that the prediction process can be expressed as follows
In the formula
And
actual values of passenger demand and random error at time t, respectively;
and
is a dieParameters and weights for the types, where p and q are positive integers representing the order of the model. Both the order and the weights of the model can be derived from the historical time series using autocorrelation functions and partial autocorrelation functions. These values can be used to detect the presence of periodicity, and the frequency of its periodicity.
Further, the sliding window based integration frame in step 3) comprises the following steps: the three prediction models proposed in the step 2) respectively predict long-term, medium-term and short-term historical data. Sliding window based consolidation frameworks aim to combine them to achieve better predictions.
A set of z models representing a given time series;
representing the set of predicted values of these models for the next time segment at time t. Integrating predicted values
Can be calculated from the following formula
Wherein
Is a model
In the time window [ t-H, t]A predicted value made within a certain time period. H is the size of the sliding window defined by the user. Because the bus data information continuously comes in the subsequent time period, the time window is continuously slid, thereby ensuring that the models are in the next H timeAnd (4) normally operating in the section. To better evaluate the accuracy of the predictions, we used the well-known mechanism of time series prediction error metric, symmetric mean percent error (sMAPE).
The passenger demand prediction system in the public transport network comprises a data storage layer and a data analysis layer, wherein the data storage layer is used for storing public transport data; the data analysis layer is used for processing the bus data stored in the data storage layer through a Poisson model changing along with time, a Poisson model changing along with the weighting time, a comprehensive autoregressive moving average model and an integration frame based on a sliding window in the data analysis layer to obtain the passenger demand in the bus network.
Further, the bus data includes five attribute values: 1) The bus state value, wherein busy represents that the number of passengers is larger than the capacity of the bus, free represents that the number of passengers is smaller than the capacity of the bus, and park represents that the bus stops at the initial or final station; 2) ID of the bus stop; 3) the time of data generation; 4) a bus license plate number; 5) the GPS data corresponds to the longitude and latitude of the location.
Further, the time-varying poisson model, the weighted time-varying poisson model and the comprehensive autoregressive moving average model can solve the problems of non-uniformity, burstiness, periodicity and the like respectively.
The invention considers two key influence factors of the public transport network and the number of passengers, thereby providing a more accurate passenger demand forecasting system and method based on the public transport network for the public transport company and the passengers. The method comprehensively considers factors such as heterogeneity, burstiness and periodicity, and provides an accurate and real-time passenger demand prediction in the public transportation network for passengers. The passenger demand prediction system and method based on the public transport network can provide a more convenient and comfortable public transport travel environment for passengers, such as reducing waiting time of the passengers and avoiding the situation that the buses are excessively crowded or loosened.
Drawings
FIG. 1 is a schematic diagram of a public transportation network based passenger demand prediction system of the present invention;
FIG. 2 is a graph comparing the results of the present invention in which passenger demand is affected by time and weather changes.
Detailed Description
The invention is described in detail below with reference to the figures and examples.
As shown in fig. 1, the system for predicting the demand of passengers in the public transportation network comprises: the system comprises a bus data layer, a data preprocessing and analyzing layer, a passenger demand forecasting layer, a knowledge base and a service layer.
The data layer is used for storing bus GPS data sources, and has the functions of collecting and sorting data and providing rich source data for the upper layer. In practical application, the appropriate storage mode can be customized according to data of different application scenes. In the present invention, the sample data is stored using an XML file.
The data preprocessing and analyzing layer is responsible for preprocessing source data in the data layer, eliminating noise, filtering redundant data irrelevant to passenger demand prediction, and performing pre-classification on bus data by using methods such as statistics and the like.
The data layer and the data preprocessing and analyzing layer are only used for obtaining the bus data source suitable for processing, and the preprocessing method is only a conventional method, which is not described herein.
The passenger demand prediction layer focuses on the realization of logic functions, is the core of the whole system and is written by using Java. The method is a set of various methods such as a Poisson model changing along with time, a Poisson model changing along with weighted time, a comprehensive autoregressive moving average model, an integration frame based on a sliding window and the like. And the different methods are divided into small modules, each of which can perform a specific demand forecast.
The knowledge base stores various rules and knowledge obtained after the demand prediction analysis. Before this, we first evaluate the results detected in the demand forecasting analysis step. After user or machine evaluation, redundant or irrelevant results may be found and should be culled. The knowledge base only keeps useful knowledge and rules which are evaluated and verified and can truly reflect the requirements of passengers.
The service layer is responsible for visually presenting the predicted result to the public transport company and the passenger, and meanwhile, some operation interfaces are provided for the public transport company and the passenger to send a query request to the passenger demand prediction layer, so that convenient and comfortable service can be better provided for the passenger. The design goal of the service layer is user-friendly, comprehensive in function, lightweight and good in compatibility.
The invention discloses a passenger demand prediction system and method based on a public transport network, which comprises the following specific steps:
1) starting from the application of an actual public transport network, providing a summary description of a passenger demand prediction method;
2) the factors such as heterogeneity, burstiness and periodicity are comprehensively considered, and three different passenger demand prediction models are respectively provided;
3) a sliding window based framework is proposed to integrate the three prediction models.
The actual public transportation network in step 1) is represented as follows: suppose a bus route contains N (N)
2) Bus stop S = { S1,S2,…,SN}. The first site is a starting site, and the last site is a terminating site. The transit of the bus between stops follows a specific route and a specific timetable. Db ={d1,d2,…,djRepresents the set of destinations for j passengers on b buses getting on at station s. The running time is divided into 4 time periods (5 am to 9 am, 9 am to 1 am, 1 am to 5 pm, and 5 pm to 9 pm) according to the shift change time of the bus. The invention aims to solve the problem of predicting the number of passengers who want to take a b-way bus at a bus stop s at the time t.
The prediction model for the passenger demand in the step 2) comprises a time-varying Poisson model, a weighted time-varying Poisson model and a comprehensive autoregressive moving average model. The time-varying poisson model comprises the following steps: the probability P (n) that a certain bus stop has n buses to stop in a given time meets Poisson distribution and is defined as
In the formula
Representing the average passenger demand for bus service over a fixed period of time, the invention
Is not constant but varies over time. Therefore we consider it as a function of time
Thereby transforming the poisson distribution to non-homogeneous.
Is defined as
Where d (t) represents weekday {1= sunday, 2= monday, … }, and h (t) is the time period to which time t belongs (for example, if every 30 minutes is a time period, time 00:31 is included in the second time period). In addition, the following two equations need to be satisfied
Wherein D is the number of time periods in a day;
is the average ratio of the one-week poisson processes;
indicates relative change on day i (e.g., saturday rates are lower than tuesday);
indicates the relative change in the ith time period on day j (e.g., peak hours);
is a discrete function representing the time-varying passenger demand at bus stop s.
The weighted time-varying poisson model comprises the following steps: the time-varying poisson model only predicts the time-dependent average passenger demand, however the passenger demand is different for each bus stop. In fact, the passenger demands of many public transportation stations are sudden and affected by many accidents, such as traffic jam and weather change. Figure 2 shows the effect of time and weather changes on passenger demand. As shown in fig. 2, the demand difference of passengers at different time periods on a certain working day is obvious, and the demand of the passengers on the bus service is the largest from 7 am to 9 am (peak time on duty) and from 4 pm to 6 pm (peak time on duty). And the demand of passengers on public transport service in rainy days is less than that in normal times, because many passengers can select a taxi or a private car and other faster travel modes in rainy days. The weighted time-varying poisson model can solve the sudden problem well. The objective is to increase the correlation of the passenger demand in the last week with the passenger demand in the previous weeks. The weight w of the correlation is calculated by the famous time series method-exponential smoothing method-which is defined as
In the formula
Is an average of the passenger demand over the past time period,
is a smoothing factor, the value of which is user-defined, and the size of which ranges from 0<
<1。
The comprehensive autoregressive moving average model comprises the following steps: both previous models assume that there is periodic regularity in the passenger's demand for public transportation services, and in fact the passenger's demand is different between different stops, different work days, and different time periods on the same day. The comprehensive autoregressive moving average model can well simulate and predict univariate time series data such as traffic flow data, short-term prediction problems and the like. This has the advantage of being able to accurately represent different types of time series, such as auto-regressive time series, moving average time series, and combinations thereof. In the synthetic autoregressive moving average model, the predicted values of variables can be viewed as linear functions of historical observations and random errors. In the invention, the time sequence is regarded as the passenger demand quantity changing along with the time on a specific bus stop s, so that the prediction process can be expressed as follows
In the formula
And
respectively the actual value of the passenger demand at time t and the random error；
And
are the parameters and weights of the model, where p and q are positive integers representing the order of the model. Both the order and the weights of the model can be derived from the historical time series using autocorrelation functions and partial autocorrelation functions. These values can be used to detect the presence of periodicity, and the frequency of its periodicity.
The sliding window-based integration framework in the step 3) comprises the following steps: the three prediction models proposed in the step 2) respectively predict long-term, medium-term and short-term historical data. Sliding window based consolidation frameworks aim to combine them to achieve better predictions.
A set of z models representing a given time series;
representing the set of predicted values of these models for the next time segment at time t. Integrating predicted values
Can be calculated from the following formula
Wherein
Is a model
In the time window [ t-H, t]A predicted value made within a certain time period. H is the size of the sliding window defined by the user. Because the bus data information continuously comes in the subsequent time period, the time window is also continuously slid, so that the models are ensured to normally operate in the next H time period. To better evaluate the accuracy of the predictions, we used the well-known mechanism of time series prediction error metric, symmetric mean percent error (sMAPE).
In summary, in the passenger demand prediction system and method based on the public transportation network, the public transportation data including 1,326 buses and 806,257 bus stations are utilized, and the public transportation information of the whole smoke platform urban area in 22 weeks is comprehensively covered; factors such as non-uniformity, burstiness and periodicity are comprehensively considered; by fitting with practical application, the problem of passenger demand prediction is analyzed on the public transport network, so that an accurate and real-time passenger demand prediction method in the public transport network can be provided for public transport companies and passengers, and the prediction accuracy rate reaches 96%.
The above examples are only for illustrating the present invention, and the structure, connection mode, etc. of the components may be changed, and all equivalent changes and modifications based on the technical solution of the present invention should not be excluded from the scope of the present invention.
Claims (5)
1. The passenger demand forecasting method based on the public transport network comprises the following specific steps:
1) starting from the application of an actual public transport network, providing a summary description of a passenger demand prediction method; the actual public transportation network in step 1) is represented as follows: suppose a certain bus line contains N, N is more than or equal to 2 bus stops S ═ S1，S2，…，SN}; the first site is a starting site, and the last site is a terminating site; the bus passing among the stops follows a specific route and a specific time schedule; db＝{d1，d2，…，djRepresents a set of destinations for j passengers on board a stop s, taking a b-way bus; dividing the running time according to the shift change time of the bus4 time periods; predicting the number of passengers to be taken on a bus with a route b at a bus stop s at time t
2) The non-uniformity, the paroxysmal and the periodic factors are comprehensively considered, and three different passenger demand prediction models are respectively provided;
the prediction model for the passenger demand in the step 2) comprises a time-varying Poisson model, a weighted time-varying Poisson model and a comprehensive autoregressive moving average model:
the time-varying poisson model comprises the following steps: the probability P (n) that a certain bus stop has n buses to stop in a given time meets Poisson distribution and is defined as
Wherein the ratio of the average demand of passengers for public transport service in a fixed time period is expressed, and the value of lambda is not constant, but changes along with time; considering it as a function of time λ (t), thereby transforming the poisson distribution to non-homogeneous; λ (t) is defined as λ (t) ═ λ0δd(t)ηd(t)，h(t)Where d (t) represents work day { 1-day of week, 2-day of week, … }, and h (t) is a time period to which time t belongs, with every 30 minutes as a time period; in addition, the following two equations need to be satisfied
And
where D is the number of time periods in the day, λ0Is the mean ratio of the Poisson processes over one week, δiDenotes the relative change on day i, ηj，iRepresents the relative change of the ith time period on the jth day, and lambda (t) is a discrete function and is used for representing the passenger demand on the bus stop s along with the change of time;
the weighted time-varying poisson model comprises the following steps: increasing the correlation of the passenger demand for the last week with the passenger demand for the previous weeks; the weight w of the correlation is calculated by a time series method, exponential smoothing, which is defined as w ═ α {1, (1- α)，(1-a)2，...，(1-a)λ-1Where λ is the average of the passenger demand during the past time period, α is a smoothing factor whose value is user defined and whose magnitude range is 0 < α < 1;
the comprehensive autoregressive moving average model comprises the following steps: simulating and predicting different types of univariate time series data, wherein the predicted values of the variables are linear functions of historical observation and random errors; the time-varying passenger demand at a particular bus stop s is considered to be time-series, so the prediction process is represented as Rs，t＝θ0+φ1Xs，t-1+φ2Xs，t-2+...+φpXs，t-p+εs，t-θ1Xs，t-1-θ2Xs，t-2-...-θqXs，t-qIn the formula, Rs，tAnd εs，tRespectively the actual value of the passenger demand at time t and a random error, phiI(1, 2.., p) and θm(1, 2.. q.) are parameters and weights of the model, where p and q are positive integers representing the order of the model; the order and the weight of the model are obtained from the historical time sequence by utilizing an autocorrelation function and a partial autocorrelation function; these values are used to detect the presence of periodicity, and the frequency of its periodicity;
3) a sliding window-based framework is provided for integrating three prediction models, comprising the following steps: the three models are combined to realize better prediction; m ═ M1，M2，...MzDenotes a set of z models modeling a given time series; mt＝{M1t，M2t，...MztRepresents the set of predicted values of these models for the next time period at time t; integration of prediction values EtCalculated from the following formula
Where ρ isjMIs a model MjIn a time window[t-H，t]A predicted value made for a time period within, H being the size of the sliding window defined by the user; because the bus data information continuously comes in the subsequent time period, the time window is also continuously slid, so that the models are ensured to normally operate in the next H time period; in order to better evaluate the accuracy of prediction, a time series prediction error metric mechanism, namely symmetric average percentage error, is adopted.
2. The method as claimed in claim 1, wherein the operation time is divided into 4 time periods from 5 am to 9 am, from 9 am to 1 am, from 1 am to 5 pm, and from 5 pm to 9 pm, according to the shift time of the bus in step 1).
3. A system adopting the passenger demand forecasting method based on the public transportation network as claimed in any one of claims 1 to 2, characterized by comprising a data storage layer and a data analysis layer, wherein the data storage layer is used for storing public transportation data; the data analysis layer is used for processing the bus data stored in the data storage layer through a Poisson model changing along with time, a Poisson model changing along with the weighting time, a comprehensive autoregressive moving average model and an integration frame based on a sliding window in the data analysis layer to obtain the passenger demand in the bus network.
4. The system of claim 3, wherein the transit data includes five attribute values: 1) the bus state value, wherein busy represents that the number of passengers is larger than the capacity of the bus, free represents that the number of passengers is smaller than the capacity of the bus, and park represents that the bus stops at the initial or final station; 2) ID of the bus stop; 3) the time of data generation; 4) a bus license plate number; 5) the GPS data corresponds to the longitude and latitude of the location.
5. The system of claim 3, wherein the time-varying Poisson model, the weighted time-varying Poisson model, and the synthetic autoregressive moving average model solve non-uniformity, burstiness, and periodicity problems, respectively.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310293407.3A CN103366224B (en) | 2013-07-15 | 2013-07-15 | Passenger demand prediction system and method based on public transport network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310293407.3A CN103366224B (en) | 2013-07-15 | 2013-07-15 | Passenger demand prediction system and method based on public transport network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103366224A CN103366224A (en) | 2013-10-23 |
| CN103366224B true CN103366224B (en) | 2021-09-17 |
Family
ID=49367517
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310293407.3A Active CN103366224B (en) | 2013-07-15 | 2013-07-15 | Passenger demand prediction system and method based on public transport network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103366224B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103955973B (en) * | 2014-04-15 | 2016-02-24 | 浙江工业大学 | A kind of detection of the bus volume of passenger traffic and acquisition methods |
| CN105489000A (en) * | 2015-09-08 | 2016-04-13 | 同济大学 | Night-shift bus stop and path selection method |
| CN106529787A (en) * | 2016-11-02 | 2017-03-22 | 南京国电南自轨道交通工程有限公司 | Subway energy management system based on face detection technology |
| TWI635447B (en) * | 2017-06-16 | 2018-09-11 | 宏碁股份有限公司 | Method and System of Predicting Passengers' Demand |
| CN108199888B (en) * | 2018-01-09 | 2021-05-04 | 东北大学 | Method for predicting and relieving public transport network congestion based on historical information |
| CN110853156B (en) * | 2019-11-18 | 2020-11-17 | 西南交通大学 | Passenger OD identification method integrating bus GPS track and IC card data |
| CN111723871B (en) * | 2020-07-09 | 2022-05-27 | 广州市公共交通数据管理中心有限公司 | Estimation method for real-time carriage full load rate of bus |
| CN113554878A (en) * | 2021-09-18 | 2021-10-26 | 深圳市城市交通规划设计研究中心股份有限公司 | Road section impedance function determination method, calculation device and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101620781A (en) * | 2008-06-30 | 2010-01-06 | 株式会社查纳位资讯情报 | System and method for forecasting passenger information and searching the same |
| CN102436603A (en) * | 2011-08-29 | 2012-05-02 | 北京航空航天大学 | Rail transit full-road-network passenger flow prediction method based on probability tree destination (D) prediction |
| CN103310287A (en) * | 2013-07-02 | 2013-09-18 | 北京航空航天大学 | Rail transit passenger flow predicting method for predicting passenger travel probability and based on support vector machine (SVM) |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100185486A1 (en) * | 2009-01-21 | 2010-07-22 | Disney Enterprises, Inc. | Determining demand associated with origin-destination pairs for bus ridership forecasting |
-
2013
- 2013-07-15 CN CN201310293407.3A patent/CN103366224B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101620781A (en) * | 2008-06-30 | 2010-01-06 | 株式会社查纳位资讯情报 | System and method for forecasting passenger information and searching the same |
| CN102436603A (en) * | 2011-08-29 | 2012-05-02 | 北京航空航天大学 | Rail transit full-road-network passenger flow prediction method based on probability tree destination (D) prediction |
| CN103310287A (en) * | 2013-07-02 | 2013-09-18 | 北京航空航天大学 | Rail transit passenger flow predicting method for predicting passenger travel probability and based on support vector machine (SVM) |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103366224A (en) | 2013-10-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103366224B (en) | Passenger demand prediction system and method based on public transport network | |
| Amirgholy et al. | Optimal design of sustainable transit systems in congested urban networks: A macroscopic approach | |
| CN110428117B (en) | Passenger flow accurate induction method and system under urban rail transit multi-scene | |
| CN102081859B (en) | Control method of bus arrival time prediction model | |
| CN107330547B (en) | Urban public transport dynamic scheduling optimization method and system | |
| CN110782070A (en) | Urban rail transit emergency passenger flow space-time distribution prediction method | |
| Moreira-Matias et al. | On predicting the taxi-passenger demand: A real-time approach | |
| Li et al. | Public bus arrival time prediction based on traffic information management system | |
| CN110675629A (en) | Big data-based highway congestion prediction and active prevention and control method | |
| Zhou et al. | The passenger demand prediction model on bus networks | |
| CN110458456B (en) | Demand response type public transportation system scheduling method and system based on artificial intelligence | |
| CN106816009B (en) | Highway real-time traffic congestion road conditions detection method and its system | |
| CN102044149A (en) | City bus operation coordinating method and device based on time variant passenger flows | |
| CN109344991B (en) | Method for predicting passenger flow of highest section of bus line | |
| CN109300326B (en) | Bus short-time dynamic scheduling method based on real-time GPS | |
| CN112309122A (en) | Intelligent bus grading decision-making system based on multi-system cooperation | |
| CN113222387A (en) | Multi-objective scheduling and collaborative optimization method for hydrogen fuel vehicle | |
| Liang et al. | Short-term passenger flow prediction in urban public transport: Kalman filtering combined k-nearest neighbor approach | |
| Hörl | Implementation of an autonomous taxi service in a multi-modal traffic simulation using MATSim | |
| Pan et al. | A self-learning algorithm for predicting bus arrival time based on historical data model | |
| CN109740823B (en) | Taxi taking decision method and system oriented to real-time scene calculation | |
| Miao et al. | Robust taxi dispatch under model uncertainties | |
| CN111723871B (en) | Estimation method for real-time carriage full load rate of bus | |
| CN112669595B (en) | Network taxi booking flow prediction method based on deep learning | |
| CN113408833A (en) | Public traffic key area identification method and device and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |