CN111145541A - Traffic flow data prediction method, storage medium, and computer device - Google Patents
Traffic flow data prediction method, storage medium, and computer device Download PDFInfo
- Publication number
- CN111145541A CN111145541A CN201911309097.3A CN201911309097A CN111145541A CN 111145541 A CN111145541 A CN 111145541A CN 201911309097 A CN201911309097 A CN 201911309097A CN 111145541 A CN111145541 A CN 111145541A
- Authority
- CN
- China
- Prior art keywords
- long
- short
- traffic
- term
- neural network
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/065—Traffic control systems for road vehicles by counting the vehicles in a section of the road or in a parking area, i.e. comparing incoming count with outgoing count
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Traffic Control Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a traffic flow data prediction method, which comprises the following steps: acquiring historical traffic flow data related to a moment to be predicted; extracting the spatial features of the historical traffic flow data by using the trained spatial feature extraction network; inputting the extracted spatial features into a trained time sequence feature extraction network, wherein the time sequence feature extraction network comprises a first long short-term memory neural network and a second long short-term memory neural network; the first long-short term memory neural network outputs a long-term timing characteristic, and the second long-short term memory neural network outputs a short-term timing characteristic and a congestion characteristic; and outputting traffic flow data at the moment to be predicted according to the long-term time sequence characteristic, the short-term time sequence characteristic and the congestion characteristic. Compared with the traditional long-short term memory neural network, the long-short term memory neural network based on the ordered neurons is more suitable for capturing the influence caused by the emergency and adapting to variable traffic environments.
Description
Technical Field
The invention belongs to the technical field of information, and particularly relates to a training method and a prediction method of a traffic flow prediction model, a computer readable storage medium and computer equipment.
Background
With the improvement of the living standard of people and the change of consumption concept, cars have become the standard configuration of each family. Meanwhile, the traffic network planning of many cities is early, and no road planning space is reserved for the development of future cities. Due to the rapid growth in the number of cars and the unreasonable planning of old city roads, traffic congestion is almost a stubborn problem in every developing city. Extreme traffic congestion at the morning and evening peaks of each day reduces people's life experience. Therefore, reasonable path planning or accurate peak shifting travel is a great demand for people in congested cities. However, these techniques all rely on accurate predictions of global traffic flow over a future period of time. Otherwise, if only the current congestion degree is relied on for path planning, and the time required by the running of the automobile is not considered, certain deviation is inevitably caused. Therefore, it is extremely important to improve the accuracy of traffic flow prediction.
Predicting traffic flow is a fundamental operation for further path planning and the like. More accurate traffic flow predictions may have better results for subsequent steps. Therefore, traffic flow prediction has been a concern for some researchers. As machine learning has evolved, more and more people are beginning to use machine learning tools to predict traffic problems. Long Short-Term Memory neural networks (LSTM) are the tools that are used to process the temporal direction of correlation of a point at most. CNN convolutional neural networks and GCN graph convolutional neural networks are the tools most used to process correlation information for spatial dimensions. Attention has also been drawn to the direction of the time sequence.
In 2019, AAAI, the Chinese team at Pennsylvania State university in America proposed a framework based on LSTM long-short term memory neural network and CNN, which achieves the best effect on the prediction of traffic flow data at that time.
In the existing technology, a common LSTM is used to process a time sequence to extract features of a long-term network and a short-term network in a time direction for training. While the general LSTM long-short term memory neural network mechanism can achieve certain effect, the structure of the network cannot be well adapted to the characteristics of traffic data. Traffic flow data, in addition to having a common long and short term dependence, the traffic flow of a node at certain times has a huge impact on the later period. For example, if a traffic accident happens in a certain place and the traffic flow in the place is seriously blocked, the traffic flow in the place will affect the traffic flow for a long time later, but the common LSTM can provide insufficient capacity for the characteristics of the node.
Disclosure of Invention
(I) technical problems to be solved by the invention
The technical problem solved by the invention is as follows: how to effectively extract the node characteristics of the emergency so as to improve the accuracy of traffic flow prediction.
(II) the technical scheme adopted by the invention
A traffic flow data prediction method, comprising:
acquiring historical traffic flow data related to a moment to be predicted;
extracting the spatial features of the historical traffic flow data by using the trained spatial feature extraction network;
inputting the extracted spatial features into a trained time sequence feature extraction network, wherein the time sequence feature extraction network comprises a first long short-term memory neural network and a second long short-term memory neural network;
the first long-short term memory neural network outputs a long-term timing characteristic, and the second long-short term memory neural network outputs a short-term timing characteristic and a congestion characteristic;
and outputting traffic flow data at the moment to be predicted according to the long-term time sequence characteristic, the short-term time sequence characteristic and the congestion characteristic.
Preferably, the historical traffic flow data comprises long-term traffic data and short-term traffic data, the long-term traffic data is traffic data in a corresponding time section m days before the time to be predicted, wherein the corresponding time section is a section containing the same time as the time to be predicted; the short-term traffic data is traffic data in a time section that is the same day as and before the time to be predicted.
Preferably, the spatial feature extraction network is a self-attention convolutional neural network, and the self-attention convolutional neural network is used for extracting long-distance spatial features and short-distance spatial features in the historical traffic flow data.
Preferably, the self-attention convolutional neural network comprises a self-attention convolutional layer for extracting distant spatial features in the historical traffic flow data.
Preferably, the traffic data in the corresponding time section of each day of the previous m days are input into each long-short term memory neural network in a one-to-one correspondence manner to generate m low-order long-term time sequence characteristics;
the m low-order long-term timing characteristics are input into a long-term short-term memory neural network to generate high-order long-term timing characteristics.
Preferably, the second long-short term memory neural network is an ordered neuron-based long-short term memory neural network, and the method for outputting the short-term timing characteristics and the congestion characteristics by the second long-short term memory neural network comprises the following steps:
identifying congestion data and non-congestion data in the short-term traffic data, wherein the congestion data represents abrupt change data generated due to an emergency in a time section before the time to be predicted;
and inputting congestion data as high-level information into a hidden layer of the second long-short term memory neural network to generate congestion characteristics, and inputting non-congestion data as low-level information into the hidden layer of the second long-short term memory neural network to generate short-term time sequence characteristics.
Preferably, the method for outputting traffic flow data of a time to be predicted according to the long-term time sequence feature, the short-term time sequence feature and the congestion feature comprises the following steps:
and fusing the long-term time sequence feature, the short-term time sequence feature and the congestion feature by using the full-connection layer to obtain a fusion feature, and converting the fusion feature into traffic flow data in a grid form through the full-connection layer.
The invention also discloses a computer readable storage medium, wherein the computer readable storage medium stores a traffic flow data prediction program, and the traffic flow data prediction program realizes the traffic flow data prediction method when being executed by a processor.
The invention also discloses a computer device which comprises a computer readable storage medium, a processor and a traffic flow data prediction program stored in the computer readable storage medium, wherein the traffic flow data prediction program realizes the traffic flow data prediction method when being executed by the processor.
(III) advantageous effects
The invention discloses a traffic flow data prediction method, which uses a long-short term memory neural network based on ordered neurons, compared with the traditional long-short term memory neural network, the long-short term memory neural network based on ordered neurons can capture the influence caused by an emergency and is more suitable for changeable traffic environments; the convolutional neural network based on the self-attention mechanism is used, and the defect that the conventional convolutional neural network is insufficient in the capability of extracting the long-distance correlation characteristics is overcome.
Drawings
Fig. 1 is a flowchart of a traffic flow data prediction method according to an embodiment of the present invention;
fig. 2 is a block diagram showing the construction of a traffic flow data prediction model according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of the structure of LSTM neurons according to an embodiment of the present invention;
FIG. 4 is a schematic structural diagram of an LSTM ordered neuron based on an embodiment of the present invention;
FIG. 5 is a graph showing experimental results of an embodiment of the present invention;
FIG. 6 is a functional block diagram of a computer device of an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The data format of the traffic flow is the grid data of the sequence, and in order to process the data in the format, a user needs to process the correlation of the information in the spatial dimension and the information in the time dimension. The traffic flow prediction model obtained through training can be used for predicting the traffic flow at a future moment, and mainly comprises a convolution neural network for extracting space characteristics and a long-term and short-term memory neural network for extracting time sequence characteristics.
As indicated by the background art, the conventional long and short term memory neural network (LSTM) in the prior art cannot effectively extract the characteristic data of some burst nodes, for example, if a traffic accident occurs in a certain place and the traffic flow in the place is seriously blocked, the traffic flow in the place will affect the traffic flow for a long time later. But the common long-short term memory neural network can not sufficiently extract the characteristics of the node. Therefore, we introduce an ordered neuron-based LSTM long-short term memory neural network model to improve the feature extraction capability for these special points.
Specifically, as shown in fig. 1, a traffic flow data prediction method disclosed in the present application includes the following steps:
step S10: historical traffic flow data related to the time to be predicted is obtained.
Specifically, the historical traffic flow data includes long-term traffic data and short-term traffic data, wherein the long-term traffic data is traffic data within a corresponding time segment m days before the time to be predicted, wherein the corresponding time segment is a segment including the same time as the time to be predicted. The short-term traffic data is traffic data in a time section that is the same day as and before the time to be predicted. For example, the traffic flow data at the time to be predicted is 12 months, 14 days and 18 days, the short-term traffic data to be acquired is 12 months, 14 days and 15 hours to 17 hours, that is, the time zone is 15 hours to 17 hours, although the size of the time zone may be set by itself, and is not limited. The long-term traffic data to be acquired are traffic flow data from 12/month 7 to 12/month 13/day from 16 hours to 20 hours, that is, traffic data of the first 7 days is acquired. The corresponding time segment is 16 hours to 20 hours, which includes the same time as the time 18 to be predicted. It should be noted that the size of the corresponding time segment may be set by itself, for example, the corresponding time segment may be centered on the time to be predicted. In addition, the value m is not limited and can be selected by oneself.
Step S20: and extracting the spatial characteristics of the historical traffic flow data by using the trained spatial characteristic extraction network.
In the prior art, a common convolutional neural network is generally adopted to extract spatial features, but the common convolutional neural network is mainly used for extracting local feature correlation, and the feature correlation of near-adjacent points is extracted in the embodiment of traffic data, but the extraction capability of the correlation between long-distance points is weak. As shown in fig. 2, the spatial feature extraction network 10 in the present application employs a self-attention convolutional neural network for this purpose, and introduces a self-attention mechanism, so that both short-distance spatial features and long-distance spatial features can be extracted. In particular, the self-attention convolutional neural network includes a self-attention convolutional layer and several general convolutional and pooling layers. The self-attention convolutional layer is arranged on the first layer, so that the spatial correlation between two points far away in the historical traffic flow data can be acquired.
Specifically, the construction method of the self-attention convolution layer mainly comprises the steps of constructing a self-attention matrix and normalizing. For the input historical data, we select the data at a time. The data is in the form of a tensor having a height H, a width W, and a dimension F, and is denoted by X ═ H, W, and F. Firstly, the dimension of an X matrix is reduced, the dimensions of H and W are reduced from two dimensions to one dimension, and the dimension is written into a form of a column through the matrix straightening operation (vec). X becomes a two-dimensional matrix of (HW, F). Constructing a self-attention matrix:
A＝XWq(WqX)T，X∈RHW×F，Wq，Wq∈RF×d
d is the dimension limit of the trainable parameter matrix to W, and is used for controlling the number of training parameters and the expressive power of self attention. Then, the self-attention moment array is normalized by performing softmax function operation. Let the resulting matrix be A. Using trainable self-attention matrices A, X and trainable matrix WvWe can build a network structure based on self-attention convolutional layers:
O＝AXWv
in the case where the correlation between nodes at a relatively long distance in the input history data, that is, the long-distance spatial feature, can be extracted by the self-attention convolution layer, it should be noted that the self-attention convolution layer of the present application can extract the correlation between nodes at a relatively long distance as well as the correlation between nodes at a relatively short distance. And the characteristics of the traffic data can be better extracted by combining with the common convolutional layer. At the end of the network structure, feature compression is performed using the full connectivity layer.
Step S30: and inputting the extracted spatial features into a trained time sequence feature extraction network, wherein the time sequence feature extraction network comprises a first long short-term memory neural network and a second long short-term memory neural network.
In particular, long short term memory neural networks (LSTM) are an improved model based on Recurrent Neural Networks (RNNs). The model established by the traditional RNN is difficult to deal with the problem of long-term dependence because gradient disappearance and gradient explosion easily occur in the training process. The basic unit of the hidden layer of LSTM is a special neuron structure and no longer a traditional neuron node. The inflow and outflow of information and the update of the previous state are respectively realized through an input gate, an output gate and a forgetting gate in the neuron structure. It is this special neuronal structure that enables LSTM to solve the problem of gradient disappearance or gradient explosion that RNNs present.
The first long-short term memory neural network 21 of the present application employs a general LSTM for processing long-term traffic data, and its schematic diagram of the neuron structure is shown in fig. 3. Wherein, each LSTM neuron structure has three gates, namely an input gate, an output gate and a forgetting gate, and the calculation formula of each gate is
ft＝σ(Wfxt+Ufht-1+bf)
it＝σ(Wixt+Uiht-1+bi)
ot＝σ(Woxt+Uoht-1+bo)
Wherein, the symbol
Expressing the multiplication of elements at corresponding positions of two vectors, wherein 'sigma' expresses a sigmoid function and is used as an activation function; f. oftRepresenting forgetting gate, itRepresenting forgetting gate, otRepresents an output gate; h ist-1Indicating the history information of the previous time, xtInformation indicating the current time, ctThe state of the neuron is represented by,
indicating the currently entered cell state.
The second long-short term memory neural network adopts LSTM based on ordered neurons for processing short-term traffic data, and the schematic diagram of the neuron structure is shown in FIG. 4. Compared with the common LSTM, the LSTM based on the ordered neurons has a neuron structure comprising five gates, namely an input gate, an output gate, a forgetting gate, a main forgetting gate and a main input gate, wherein the calculation formula of each gate is
ft＝σ(Wfxt+Ufht-1+bf)
it＝σ(Wixt+Uiht-1+bi)
ot＝σ(Woxt+Uoht-1+bo)
Symbol
Expressing the multiplication of elements at corresponding positions of two vectors, wherein 'sigma' expresses a sigmoid function and is used as an activation function; f. oftRepresenting forgetting gate, itRepresenting forgetting gate, otWhich represents the output gate or gates, respectively,
the main forgetting door is shown to be,
represents a master input gate; h ist-1Indicating the history information of the previous time, xtInformation indicating the current time, ctRepresenting neuronal state，
Indicating the currently entered cell state.
Indicating a cumsum operation to the right,
indicating a cumsum operation to the left.
The LSTM based on the ordered neurons has the core idea that information of high and low levels is distinguished and updated in a partitioned mode, after the neurons are ordered, the height of the information level is represented through the front and the back of the position, then when the neurons are updated, the level of historical information at the previous moment and the level of current input information are predicted respectively, and the neurons are updated in a partitioned mode through the two levels. Put another way, higher level information may remain a significant distance because the higher level copies the history information directly, resulting in the history information being copied and unchanged, while lower level information may be updated at each input step, embedding the hierarchy by information hierarchy because the lower level copies the input directly, with the input being changed. More generally, with group updates, higher group information is propagated further, i.e., spans are larger, lower group spans are smaller, and these different spans form the hierarchy of the input sequence. The method and the device utilize the characteristic of hierarchical updating to enable data characteristics which have important influence on subsequent traffic in short-term traffic data to be stored for a longer time so as to predict traffic flow data more accurately, and the specific steps are described below.
Step S40: the first long-short term memory neural network outputs long-term timing characteristics, and the second long-short term memory neural network outputs short-term timing characteristics and congestion characteristics.
Specifically, the first long-short term memory neural network comprises a plurality of long-short term memory neural networks, and the specific method for outputting the long-term timing characteristics by the first long-short term memory neural network comprises the following steps:
and the traffic data in the corresponding time section of each day of the previous m days are input into each long-short term memory neural network in a one-to-one correspondence manner to generate m low-order long-term time sequence characteristics. An attention matrix is set in the part, and weighted summation is carried out through calculation and similarity of predicted values in the short-term neural network. Then, the m low-order long-term timing characteristics are inputted into a long-term short-term memory neural network to generate high-order long-term timing characteristics.
Further, the method for outputting the short-term timing characteristics and the congestion characteristics by the second long-short-term memory neural network comprises the following steps:
identifying congestion data and non-congestion data in the short-term traffic data, wherein the congestion data represents abrupt change data generated due to an emergency in a time section before the time to be predicted, for example, the emergency such as a traffic accident occurs;
and inputting congestion data as high-level information into a hidden layer of the second long-short term memory neural network to generate congestion characteristics, and inputting non-congestion data as low-level information into the hidden layer of the second long-short term memory neural network to generate short-term time sequence characteristics. According to the analysis, the congestion data can have important influence on the traffic at the subsequent time, so that the congestion data needs to be stored for a longer time and cannot be updated easily, and the non-congestion data can be used as low-level information to be updated normally, so that the method can adapt to more variable actual traffic environments.
It should be noted that, in general, the congestion data existing in the long-term traffic data does not significantly affect the traffic at the time to be predicted, and for example, a traffic accident occurring 3 days before the time to be predicted does not generally affect the traffic at the time to be predicted, so that the normal LSTM processing may be adopted. The congestion data existing in the short-term traffic data can obviously affect the time to be predicted, for example, traffic accidents occurring in the first two hours of the time to be predicted generally affect the traffic at the time to be predicted, so that the congestion data needs to be extracted by using the LSTM based on the ordered neurons in the application, and can be spread for a longer time, so that the traffic condition can be predicted more accurately.
Step S50: and outputting traffic flow data at the moment to be predicted according to the long-term time sequence characteristic, the short-term time sequence characteristic and the congestion characteristic.
And fusing the long-term time sequence feature, the short-term time sequence feature and the congestion feature by using the full-connection layer to obtain a fusion feature, and converting the fusion feature into traffic flow data in a grid form through the full-connection layer.
In order to prove that the prediction method of the present application has a more accurate prediction effect compared to the existing prediction method, the following experiment is performed:
in order to verify the effect of the traffic flow data prediction method, namely to evaluate the performance of the traffic flow prediction model, a large data set, taxi data in New York City, is selected, the evaluation index is the Mean Absolute Percentage Error (MAPE), and the smaller the mean absolute percentage error is, the higher the prediction precision of the representative model is. Comparative examples include the mainstream models presently in the industry for traffic prediction, ARIMA, XGboost, ConvLSTM and STDN. The experimental results are shown in fig. 5, where ST-AOCNet denotes the traffic flow prediction model of the present application, Start denotes the departure from the grid at this time, and End denotes the arrival at the grid at this time. The experimental data are calculated according to a time interval, data are collected every half hour, Start refers to data leaving the grid within half hour, and End refers to data arriving at the grid within half hour. According to experimental results, the MAPE value of the traffic flow prediction model is smaller than that of the existing main flow model, namely the accuracy of the applied traffic flow prediction model is higher than that of the existing main flow model.
According to experimental results, the traffic flow data prediction method disclosed by the application uses the long-short term memory neural network based on the ordered neurons, and compared with the traditional long-short term memory neural network and the long-short term memory neural network based on the ordered neurons, the long-short term memory neural network based on the ordered neurons can capture the influence caused by the emergency events and is more suitable for variable traffic environments; the convolutional neural network based on the self-attention mechanism is used, and the defect that the conventional convolutional neural network is insufficient in the capability of extracting the long-distance correlation characteristics is overcome.
The application also discloses a computer readable storage medium, wherein a traffic flow data prediction program is stored in the computer readable storage medium, and the traffic flow data prediction program realizes the traffic flow data prediction method when being executed by a processor.
The present application also discloses a computer device, and on the hardware level, as shown in fig. 6, the terminal includes a processor 12, an internal bus 13, a network interface 14, and a computer-readable storage medium 11. The processor 12 reads a corresponding computer program from the computer-readable storage medium and then runs, forming a request processing apparatus on a logical level. Of course, besides software implementation, the one or more embodiments in this specification do not exclude other implementations, such as logic devices or combinations of software and hardware, and so on, that is, the execution subject of the following processing flow is not limited to each logic unit, and may also be hardware or logic devices. The computer-readable storage medium 11 has stored thereon a traffic flow data prediction program that, when executed by a processor, implements the traffic flow data prediction method described above.
Computer-readable storage media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer-readable storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic disk storage, quantum memory, graphene-based storage media or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device.
Although a few embodiments of the present invention have been shown and described, it would be appreciated by those skilled in the art that changes may be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the claims and their equivalents, and that such changes and modifications are intended to be within the scope of the invention.
Claims (9)
1. A traffic flow data prediction method characterized by comprising:
acquiring historical traffic flow data related to a moment to be predicted;
extracting the spatial features of the historical traffic flow data by using the trained spatial feature extraction network;
inputting the extracted spatial features into a trained time sequence feature extraction network, wherein the time sequence feature extraction network comprises a first long short-term memory neural network and a second long short-term memory neural network;
the first long-short term memory neural network outputs a long-term timing characteristic, and the second long-short term memory neural network outputs a short-term timing characteristic and a congestion characteristic;
and outputting traffic flow data at the moment to be predicted according to the long-term time sequence characteristic, the short-term time sequence characteristic and the congestion characteristic.
2. The traffic-flow data prediction method according to claim 1, characterized in that the historical traffic-flow data includes long-term traffic data and short-term traffic data, the long-term traffic data being traffic data within a corresponding time segment m days before the time to be predicted, wherein the corresponding time segment is a segment containing the same time as the time to be predicted; the short-term traffic data is traffic data in a time section that is the same day as and before the time to be predicted.
3. The traffic-flow data prediction method according to claim 2, characterized in that the spatial feature extraction network is a self-attention convolutional neural network for extracting long-distance spatial features and short-distance spatial features in the historical traffic-flow data.
4. The traffic-flow data prediction method according to claim 3, characterized in that the self-attention convolutional neural network includes a self-attention convolutional layer for extracting a distant spatial feature in the historical traffic-flow data.
5. The traffic flow data prediction method according to claim 2, wherein the first long-short term memory neural network includes a plurality of long-short term memory neural networks, and the specific method for outputting the long-short term timing characteristics by the first long-short term memory neural network includes:
inputting the traffic data in the corresponding time section of each day of the previous m days into each long-short term memory neural network in a one-to-one correspondence manner to generate m low-order long-term time sequence characteristics;
the m low-order long-term timing characteristics are input into a long-term short-term memory neural network to generate high-order long-term timing characteristics.
6. The traffic-flow data prediction method according to claim 2, wherein the second long-short term memory neural network is an ordered neuron-based long-short term memory neural network, and the method for outputting the short-term timing characteristics and the congestion characteristics by the second long-short term memory neural network includes:
identifying congestion data and non-congestion data in the short-term traffic data, wherein the congestion data represents abrupt change data generated due to an emergency in a time section before the time to be predicted;
and inputting congestion data as high-level information into a hidden layer of the second long-short term memory neural network to generate congestion characteristics, and inputting non-congestion data as low-level information into the hidden layer of the second long-short term memory neural network to generate short-term time sequence characteristics.
7. The traffic-flow data prediction method according to claim 1, wherein the method of outputting the traffic-flow data of the time to be predicted based on the long-term timing characteristics, the short-term timing characteristics, and the congestion characteristics includes:
and fusing the long-term time sequence feature, the short-term time sequence feature and the congestion feature by using the full-connection layer to obtain a fusion feature, and converting the fusion feature into traffic flow data in a grid form through the full-connection layer.
8. A computer-readable storage medium characterized in that a traffic-flow data prediction program is stored, which when executed by a processor implements the traffic-flow data prediction method according to any one of claims 1 to 7.
9. A computer device characterized by comprising a computer-readable storage medium, a processor, and a traffic-flow data prediction program stored in the computer-readable storage medium, which when executed by the processor implements the traffic-flow data prediction method according to claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911309097.3A CN111145541B (en) | 2019-12-18 | 2019-12-18 | Traffic flow data prediction method, storage medium, and computer device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911309097.3A CN111145541B (en) | 2019-12-18 | 2019-12-18 | Traffic flow data prediction method, storage medium, and computer device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111145541A true CN111145541A (en) | 2020-05-12 |
| CN111145541B CN111145541B (en) | 2021-10-22 |
Family
ID=70518715
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911309097.3A Active CN111145541B (en) | 2019-12-18 | 2019-12-18 | Traffic flow data prediction method, storage medium, and computer device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111145541B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111583648A (en) * | 2020-05-14 | 2020-08-25 | 公安部交通管理科学研究所 | Urban traffic flow prediction method fusing GPS data and checkpoint flow data |
| CN111709786A (en) * | 2020-06-18 | 2020-09-25 | 北京字节跳动网络技术有限公司 | Method, apparatus, device and medium for generating user retention time |
| CN112712695A (en) * | 2020-12-30 | 2021-04-27 | 桂林电子科技大学 | Traffic flow prediction method, device and storage medium |
| CN112966871A (en) * | 2021-03-15 | 2021-06-15 | 上海交通大学 | Traffic jam prediction method and system based on convolution long-short term memory neural network |
| CN113192325A (en) * | 2021-04-19 | 2021-07-30 | 银江股份有限公司 | Traffic incident correlation analysis method and system |
| WO2022104504A1 (en) * | 2020-11-17 | 2022-05-27 | 中国科学院深圳先进技术研究院 | Road traffic state prediction method and system, terminal, and storage medium |
| CN115359444A (en) * | 2022-10-18 | 2022-11-18 | 智道网联科技(北京)有限公司 | Road congestion prediction method and device |
| CN115359662A (en) * | 2022-10-18 | 2022-11-18 | 智道网联科技(北京)有限公司 | Lane congestion prediction method and device |
| CN115423048A (en) * | 2022-11-03 | 2022-12-02 | 山东大学 | Traffic flow abnormity detection method and system based on mode similarity |
| CN116090669A (en) * | 2023-04-03 | 2023-05-09 | 江西锦路科技开发有限公司 | Traffic flow prediction method, equipment and medium based on hybrid neural network |
| CN116109011A (en) * | 2023-04-10 | 2023-05-12 | 知鱼智联科技股份有限公司 | Energy consumption management method and terminal for intelligent park |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2002367077A (en) * | 2001-06-07 | 2002-12-20 | Mitsubishi Electric Corp | Device and method for deciding traffic congestion |
| CN102184638A (en) * | 2011-04-28 | 2011-09-14 | 北京市劳动保护科学研究所 | Method for preprocessing pedestrian traffic data |
| CN106886846A (en) * | 2017-04-26 | 2017-06-23 | 中南大学 | A kind of bank outlets' excess reserve Forecasting Methodology that Recognition with Recurrent Neural Network is remembered based on shot and long term |
| US20180137412A1 (en) * | 2016-11-16 | 2018-05-17 | Cisco Technology, Inc. | Network traffic prediction using long short term memory neural networks |
| CN108734614A (en) * | 2017-04-13 | 2018-11-02 | 腾讯科技（深圳）有限公司 | Traffic congestion prediction technique and device, storage medium |
| CN109284870A (en) * | 2018-10-08 | 2019-01-29 | 南昌大学 | Short-term method for forecasting photovoltaic power generation quantity based on shot and long term Memory Neural Networks |
| CN109389091A (en) * | 2018-10-22 | 2019-02-26 | 重庆邮电大学 | The character identification system and method combined based on neural network and attention mechanism |
| CN109547248A (en) * | 2018-11-20 | 2019-03-29 | 上海航天测控通信研究所 | Based on artificial intelligence in orbit aerocraft ad hoc network method for diagnosing faults and device |
| CN109754605A (en) * | 2019-02-27 | 2019-05-14 | 中南大学 | A kind of traffic forecast method based on attention temporal diagram convolutional network |
| CN109886387A (en) * | 2019-01-07 | 2019-06-14 | 北京大学 | It is a kind of that the traffic time sequence forecasting method returned is promoted based on gating network and gradient |
| CN109920250A (en) * | 2019-05-10 | 2019-06-21 | 李天学 | Dynamic prediction urban road intelligent traffic administration system method |
| KR102021992B1 (en) * | 2018-08-21 | 2019-09-18 | 한국과학기술정보연구원 | Apparatus for controling a trafic signal, method for controling a trafic signal, and recoding medium for controling a tarfic signal |
| CN110580287A (en) * | 2019-08-20 | 2019-12-17 | 北京亚鸿世纪科技发展有限公司 | Emotion classification method based ON transfer learning and ON-LSTM |
-
2019
- 2019-12-18 CN CN201911309097.3A patent/CN111145541B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2002367077A (en) * | 2001-06-07 | 2002-12-20 | Mitsubishi Electric Corp | Device and method for deciding traffic congestion |
| CN102184638A (en) * | 2011-04-28 | 2011-09-14 | 北京市劳动保护科学研究所 | Method for preprocessing pedestrian traffic data |
| US20180137412A1 (en) * | 2016-11-16 | 2018-05-17 | Cisco Technology, Inc. | Network traffic prediction using long short term memory neural networks |
| CN108734614A (en) * | 2017-04-13 | 2018-11-02 | 腾讯科技（深圳）有限公司 | Traffic congestion prediction technique and device, storage medium |
| CN106886846A (en) * | 2017-04-26 | 2017-06-23 | 中南大学 | A kind of bank outlets' excess reserve Forecasting Methodology that Recognition with Recurrent Neural Network is remembered based on shot and long term |
| KR102021992B1 (en) * | 2018-08-21 | 2019-09-18 | 한국과학기술정보연구원 | Apparatus for controling a trafic signal, method for controling a trafic signal, and recoding medium for controling a tarfic signal |
| CN109284870A (en) * | 2018-10-08 | 2019-01-29 | 南昌大学 | Short-term method for forecasting photovoltaic power generation quantity based on shot and long term Memory Neural Networks |
| CN109389091A (en) * | 2018-10-22 | 2019-02-26 | 重庆邮电大学 | The character identification system and method combined based on neural network and attention mechanism |
| CN109547248A (en) * | 2018-11-20 | 2019-03-29 | 上海航天测控通信研究所 | Based on artificial intelligence in orbit aerocraft ad hoc network method for diagnosing faults and device |
| CN109886387A (en) * | 2019-01-07 | 2019-06-14 | 北京大学 | It is a kind of that the traffic time sequence forecasting method returned is promoted based on gating network and gradient |
| CN109754605A (en) * | 2019-02-27 | 2019-05-14 | 中南大学 | A kind of traffic forecast method based on attention temporal diagram convolutional network |
| CN109920250A (en) * | 2019-05-10 | 2019-06-21 | 李天学 | Dynamic prediction urban road intelligent traffic administration system method |
| CN110580287A (en) * | 2019-08-20 | 2019-12-17 | 北京亚鸿世纪科技发展有限公司 | Emotion classification method based ON transfer learning and ON-LSTM |
Non-Patent Citations (3)
| Title |
|---|
| SOONER高: "《CSDN ON-lstm：用有序神经元表达层次结构》", 20 August 2019 * |
| YIKANG SHEN 等: "Ordered Neurons：Integrating Tree Structures into Recurrent Neural Network", 《ARXIV》 * |
| 陈韫: "基于LSTM深度网络的城市道路短时交通状态预测模型研究", 《中国优秀硕士学位论文全文数据库 工程科技II辑》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111583648A (en) * | 2020-05-14 | 2020-08-25 | 公安部交通管理科学研究所 | Urban traffic flow prediction method fusing GPS data and checkpoint flow data |
| CN111709786A (en) * | 2020-06-18 | 2020-09-25 | 北京字节跳动网络技术有限公司 | Method, apparatus, device and medium for generating user retention time |
| CN111709786B (en) * | 2020-06-18 | 2024-04-30 | 抖音视界有限公司 | Method, apparatus, device and medium for generating user retention time |
| WO2022104504A1 (en) * | 2020-11-17 | 2022-05-27 | 中国科学院深圳先进技术研究院 | Road traffic state prediction method and system, terminal, and storage medium |
| CN112712695A (en) * | 2020-12-30 | 2021-04-27 | 桂林电子科技大学 | Traffic flow prediction method, device and storage medium |
| CN112712695B (en) * | 2020-12-30 | 2021-11-26 | 桂林电子科技大学 | Traffic flow prediction method, device and storage medium |
| CN112966871B (en) * | 2021-03-15 | 2023-05-05 | 上海交通大学 | Traffic jam prediction method and system based on convolution long-short-term memory neural network |
| CN112966871A (en) * | 2021-03-15 | 2021-06-15 | 上海交通大学 | Traffic jam prediction method and system based on convolution long-short term memory neural network |
| CN113192325A (en) * | 2021-04-19 | 2021-07-30 | 银江股份有限公司 | Traffic incident correlation analysis method and system |
| CN113192325B (en) * | 2021-04-19 | 2022-06-24 | 银江技术股份有限公司 | Traffic incident correlation analysis method and system |
| CN115359662A (en) * | 2022-10-18 | 2022-11-18 | 智道网联科技(北京)有限公司 | Lane congestion prediction method and device |
| CN115359662B (en) * | 2022-10-18 | 2023-01-10 | 智道网联科技(北京)有限公司 | Lane congestion prediction method and device |
| CN115359444A (en) * | 2022-10-18 | 2022-11-18 | 智道网联科技(北京)有限公司 | Road congestion prediction method and device |
| CN115423048A (en) * | 2022-11-03 | 2022-12-02 | 山东大学 | Traffic flow abnormity detection method and system based on mode similarity |
| CN116090669A (en) * | 2023-04-03 | 2023-05-09 | 江西锦路科技开发有限公司 | Traffic flow prediction method, equipment and medium based on hybrid neural network |
| CN116109011A (en) * | 2023-04-10 | 2023-05-12 | 知鱼智联科技股份有限公司 | Energy consumption management method and terminal for intelligent park |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111145541B (en) | 2021-10-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111145541B (en) | Traffic flow data prediction method, storage medium, and computer device | |
| CN109697852B (en) | Urban road congestion degree prediction method based on time sequence traffic events | |
| Fang et al. | Attention meets long short-term memory: A deep learning network for traffic flow forecasting | |
| CN115240425B (en) | Traffic prediction method based on multi-scale space-time fusion graph network | |
| CN109840660A (en) | A kind of vehicular characteristics data processing method and vehicle risk prediction model training method | |
| CN111612243A (en) | Traffic speed prediction method, system and storage medium | |
| CN114944053B (en) | Traffic flow prediction method based on space-time hypergraph neural network | |
| EP4016412A1 (en) | Traffic prediction | |
| CN111915081B (en) | Peak sensitive travel demand prediction method based on deep learning | |
| CN110570035B (en) | People flow prediction system for simultaneously modeling space-time dependency and daily flow dependency | |
| CN112633602B (en) | Traffic congestion index prediction method and device based on GIS map information | |
| CN114360239A (en) | Traffic prediction method and system for multilayer space-time traffic knowledge map reconstruction | |
| CN113780665A (en) | Private car stop position prediction method and system based on enhanced recurrent neural network | |
| CN114565187A (en) | Traffic network data prediction method based on graph space-time self-coding network | |
| CN116311939A (en) | OD demand prediction method and system based on dynamic space-time correlation | |
| CN116523104A (en) | Abnormal group flow prediction method and device based on context awareness and deep learning | |
| CN115099328A (en) | Traffic flow prediction method, system, device and storage medium based on countermeasure network | |
| Xing et al. | RL-GCN: Traffic flow prediction based on graph convolution and reinforcement learning for smart cities | |
| CN114781696B (en) | Model-free accident influence range prediction method for urban road network | |
| CN114566048B (en) | Traffic control method based on multi-view self-adaptive space-time diagram network | |
| CN116612633A (en) | Self-adaptive dynamic path planning method based on vehicle-road cooperative sensing | |
| CN113971496A (en) | Urban traffic network state evolution trend prediction method and system under influence of activities | |
| Lu et al. | Graph hierarchical convolutional recurrent neural network (GHCRNN) for vehicle condition prediction | |
| Ma et al. | Data Model of Key Indicators of Urban Architectural Design Based on Long‐and Short‐Term Convolutional Memory Network | |
| Yang | [Retracted] Urban Traffic Flow Prediction with Deep Neural Network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |