CN115222089A - Road traffic jam prediction method, device, equipment and readable storage medium - Google Patents
Road traffic jam prediction method, device, equipment and readable storage medium Download PDFInfo
- Publication number
- CN115222089A CN115222089A CN202210604273.1A CN202210604273A CN115222089A CN 115222089 A CN115222089 A CN 115222089A CN 202210604273 A CN202210604273 A CN 202210604273A CN 115222089 A CN115222089 A CN 115222089A
- Authority
- CN
- China
- Prior art keywords
- road
- data
- traffic
- congestion
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 239000011159 matrix material Substances 0.000 claims abstract description 60
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 32
- 230000007246 mechanism Effects 0.000 claims abstract description 24
- 238000012545 processing Methods 0.000 claims abstract description 18
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 17
- 238000004364 calculation method Methods 0.000 claims description 35
- 238000013528 artificial neural network Methods 0.000 claims description 22
- 230000006870 function Effects 0.000 claims description 20
- 230000000737 periodic effect Effects 0.000 claims description 13
- 238000004590 computer program Methods 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 5
- 238000001514 detection method Methods 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 3
- 238000012549 training Methods 0.000 abstract description 7
- 238000011176 pooling Methods 0.000 description 10
- 238000007477 logistic regression Methods 0.000 description 9
- 238000004891 communication Methods 0.000 description 8
- 230000004927 fusion Effects 0.000 description 8
- 238000012163 sequencing technique Methods 0.000 description 6
- 230000005236 sound signal Effects 0.000 description 4
- 230000009466 transformation Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000005065 mining Methods 0.000 description 3
- 238000013519 translation Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 230000002123 temporal effect Effects 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000009412 basement excavation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 210000000653 nervous system Anatomy 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000000844 transformation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/40—Business processes related to the transportation industry
Landscapes
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Game Theory and Decision Science (AREA)
- Educational Administration (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention provides a road traffic jam prediction method, a device, equipment and a readable storage medium, which relate to the technical field of road jam prediction and comprise the steps of calculating matrix data of traffic jam indexes of each road in different time periods according to first data; inputting at least one matrix data into a preset prediction model based on a convolutional neural network algorithm, and extracting at least one spatial feature data; according to a position coding strategy, carrying out position coding processing on the obtained spatial characteristic data; and inputting the processed position encoding data into a Transformer model for prediction based on an attention mechanism to obtain a road traffic jam prediction result. The method has the advantages that the training speed of the model is greatly improved, meanwhile, the hidden danger of gradient problem existing in the convolutional neural network algorithm is overcome, the accuracy of the model is further improved, and the prediction accuracy is further improved.
Description
Technical Field
The invention relates to the technical field of road congestion prediction, in particular to a road traffic congestion prediction method, a road traffic congestion prediction device, road traffic congestion prediction equipment and a readable storage medium.
Background
Urban traffic prediction is always a hotspot of research of scientific researchers, and different prediction methods are adopted in different periods, so that excellent success is achieved. However, the autoregressive moving average model usually can only deal with simple problems, and cannot be completely matched with a potential objective function in practice, so that under-fitting is easy to occur. Although the autoregressive moving average model can achieve the purpose of predicting the traffic flow, the autoregressive moving average model is only suitable for the road section with stable traffic condition because the autoregressive moving average model requires stable input data and the prediction precision is greatly reduced when the autoregressive moving average model is used for complex traffic flow data. Meanwhile, the autoregressive moving average model is difficult to capture the characteristics of the space-time correlation in traffic flow data, and cannot meet the requirement of more accurately predicting the traffic flow.
Although the artificial neural network has many advantages, the defects are further exposed, for example, the data-driven mechanism of the artificial neural network cannot well explain the spatial correlation of urban areas, and the artificial neural network has a shallow structure and low prediction precision.
Disclosure of Invention
The present invention aims to provide a road traffic congestion prediction method, apparatus, device and readable storage medium to improve the above problems. In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
in a first aspect, the present application provides a method for predicting road traffic congestion, comprising:
acquiring first data, wherein the first data are road condition data of each road in a road traffic network in different time periods;
calculating matrix data of traffic jam indexes of each road in different time periods according to the first data;
inputting at least one matrix data into a preset prediction model based on a convolutional neural network algorithm, and extracting at least one spatial feature data, wherein the spatial feature data comprises congestion factors of each road;
according to a position coding strategy, carrying out position coding processing on the obtained spatial feature data, wherein the position coding strategy comprises local position coding, global position coding and periodic position coding;
and inputting the processed position encoding data into a Transformer model for prediction based on an attention mechanism to obtain a road traffic jam prediction result.
In a second aspect, the present application further provides a road traffic congestion prediction device, including an acquisition module, a calculation module, an extraction module, a processing module, and a prediction module, wherein:
the acquisition module is used for acquiring first data, wherein the first data are road condition data of each road in a road traffic network in different time periods;
the calculation module is used for calculating matrix data of traffic jam indexes of each road in different time periods according to the first data;
the extraction module is used for inputting at least one matrix data into a preset prediction model based on a convolutional neural network algorithm and extracting at least one spatial characteristic data, wherein the spatial characteristic data comprises congestion factors of each road;
the processing module is used for carrying out position coding processing on the obtained spatial characteristic data according to a position coding strategy, wherein the position coding strategy comprises local position coding, global position coding and periodic position coding;
and the prediction module is used for inputting the processed position encoding data into a Transformer model for prediction based on an attention mechanism to obtain a road traffic jam prediction result.
In a third aspect, the present application further provides a device for predicting road traffic congestion, comprising:
a memory for storing a computer program;
a processor for implementing the steps of the road traffic congestion prediction method when executing the computer program.
In a fourth aspect, the present application further provides a readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the steps of the above-mentioned method for predicting road traffic congestion.
The beneficial effects of the invention are as follows: the convolutional neural network is combined with a Transformer to predict the urban traffic, so that the training speed of the model can be greatly improved, the hidden danger of the gradient problem of the CNN-LSTM is overcome, and the accuracy of the model is further improved; the position coding component is added, so that the position coding can be better fused into the input sequence, information about positions is added into the input sequence, and the position coding can be ensured not to hide other relations among the sequences, so that the model training is more difficult. And a position combination mode based on an attention mechanism is added to a convolutional neural network space component and a Transformer time component, and the prediction accuracy is improved by further improving the convolutional neural network combined with a Transformer model.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by the practice of the embodiments of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.
Fig. 1 is a schematic flowchart of a road traffic congestion prediction method according to an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of a road traffic congestion prediction apparatus according to an embodiment of the present invention;
fig. 3 is a schematic structural diagram of a road traffic congestion prediction device according to an embodiment of the present invention.
In the figure, 701 an acquisition module; 7011. dividing the cells; 7012. a first acquisition unit; 7013. a matching unit; 7014. a first calculation unit; 7015. a second calculation unit; 702. a calculation module; 7021. a second acquisition unit; 7022. a third calculation unit; 7023. a construction unit; 7024. a fourth calculation unit; 7025. a solving unit; 7026. a fifth calculation unit; 703. an extraction module; 7031. a setting unit; 7032. a first convolution unit; 7033. an excavation unit; 7034. a second convolution unit; 7035. a building unit; 7036. an extraction unit; 704. a processing module; 7041. a first sorting unit; 7042. a second sorting unit; 7043. a sixth calculation unit; 7044. a seventh calculation unit; 7045. an eighth calculation unit; 7046. a fusion unit; 705. a prediction module; 7051. a ninth calculation unit; 7052. a tenth calculation unit; 7053. a first input unit; 7054. a second input unit; 7055. a third input unit; 7056. a prediction unit; 800. a road traffic congestion prediction device; 801. a processor; 802. a memory; 803. a multimedia component; 804. an input/output (I/O) interface; 805. a communication component.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be obtained by a person skilled in the art without inventive step based on the embodiments of the present invention, are within the scope of protection of the present invention.
It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined or explained in subsequent figures. Meanwhile, in the description of the present invention, the terms "first", "second", and the like are used only for distinguishing the description, and are not to be construed as indicating or implying relative importance.
Example 1:
the embodiment provides a road traffic jam prediction method.
Referring to fig. 1, it is shown that the method includes step S100, step S200, step S300, step S400 and step S500.
S100, collecting first data, wherein the first data are road condition data of each road in a road traffic network in different time periods.
It is understood that step S100 includes steps S101, S102, S103, S104, and S105, where:
s101, dividing road traffic grid areas of a city to obtain road network data of the city;
s102, acquiring first positioning data, wherein the first positioning data are original positioning data generated by a driver in the process of using mobile phone navigation software, and the original positioning data comprise field information of time, longitude and latitude, speed and direction angle;
s103, matching the first positioning data to road network data of a city to obtain a positioning track matched with a road network;
s104, calculating the traffic state of each road in a fixed time period based on the positioning track;
and S105, calculating the traffic state of each divided area in the road traffic grid of the city according to the traffic state.
It should be noted that the first positioning data may be obtained from a GPS track of the taxi, the urban grid area is divided, the traffic state of each road in a fixed time period is calculated based on the GPS track of the taxi, and the traffic state of each divided area is calculated according to the traffic state of each road.
The first step is as follows: the urban grid is divided firstly, for example, the urban district is divided into 8x8 non-overlapping grid areas by a grid division method, and each grid is 3.2km long and 2.9km wide.
The second step is that: and calculating the regional traffic congestion index. The invention adopts a traffic jam index based on road speed. The INRIX Index is a typical road speed-based traffic congestion Index, which is in fact widely used in most countries in europe. It should be noted that the core of calculating the traffic congestion index is to calculate the traffic congestion index of different areas in a city according to the traffic congestion index based on the road speed.
And S200, calculating matrix data of the traffic jam indexes of each road in different time periods according to the first data.
It is understood that step S200 includes steps S201, S202, S203, S204, S205 and S206, where:
s201, obtaining the traffic flow and the average running speed of vehicles of each road section in a road network according to data provided by road network vehicle detection equipment;
s202, calculating the number of vehicles contained in each road section in the road network according to the traffic flow of each road section in the road network and the average driving speed of the vehicles;
s203, constructing a congestion intensity function with the speed as an independent variable;
s204, inputting the speed of each road section into a congestion intensity function, and calculating to obtain the average congestion intensity of the road section;
s205, using the number of vehicles of each road section as a weight of the road section, and solving a weighted average value of the congestion intensity of the whole road network according to the average congestion intensity;
and S206, taking the weighted average value as a traffic jam index of the road network, collecting road section length information, and calculating to obtain matrix data of the traffic jam index.
Note that, the calculation in step S200 is:
1. the actual link speed CSij of the link is calculated.
In this embodiment, the time interval is set to 15 minutes, so the CSij is calculated by the following formula:
in the formula: CSij represents the actual link speed for the jth time interval of link i, n represents the number of vehicles passing through link i during the jth time interval, and Sk represents the actual speed of the passing vehicles.
2. And calculating an INRIX Index value Aij of each road section in the time interval. Therefore, the calculation formula of the INRIX Index value Aij is as follows:
in the formula: RSij represents the link free flow speed for the jth time interval of the link i.
3. And calculating INRIX Index values Bij of different areas of the city. Taking the length of the road section as a weight coefficient to carry out weighting and averaging on Aij, wherein the Bij calculation formula is as follows:
in the formula: bij represents the Index value of the jth time interval INRIX Index of the urban area i. Lk represents a link length. N represents the total number of links in different areas of the city.
S300, inputting at least one matrix data into a preset prediction model based on a convolutional neural network algorithm, and extracting at least one spatial feature data, wherein the spatial feature data comprises congestion factors of each road.
It is understood that step S300 includes steps S301, S302, S303, S304, S305, and S306, where:
s301, setting matrix data as a two-dimensional matrix of 8X 8;
s302, performing first convolution on the 8X8 two-dimensional matrix based on a convolutional neural network algorithm;
s303, excavating the two-dimensional matrix after the first convolution operation to obtain a 4X4 two-dimensional matrix;
s304, performing second convolution on the 4X4 two-dimensional matrix to obtain a 2X2 two-dimensional matrix;
s305, constructing a position coding model;
and S306, based on a convolutional neural network algorithm, inputting the 2X2 two-dimensional matrix into the position coding model after passing through the fully-connected neural network, and extracting spatial characteristic data of at least one current traffic jam nearby area.
It should be noted that, this step is divided into two steps, which are respectively defined for urban area traffic prediction problem and extracted for the convolutional neural network module, wherein:
under the condition of a given time slice t, the urban traffic jam state { Yn +1, \ 8230;, yn + m } under the time slice n x t + m x t is predicted through the historical traffic jam states { Y1, Y2 \8230;, yn }, of n time slices.
For the prediction of the congestion state of urban traffic, not only the continuity and periodicity of the temporal features but also the dependency of the spatial features need to be considered. Time slice t is defined herein as 15 minutes.
The CNN is a short for convolutional neural network, is widely applied to the field of image processing, and has strong characteristic learning capability. The CNN is generated by enlightening the human visual nervous system, and the receptive field of the visual nerve is replaced by a convolution kernel, so that the calculation amount can be reduced and the characteristics of the image can be kept.
Preferably, the spatial characteristics in the invention comprise a surrounding area congestion factor, and the factor comprises the number of vehicles, namely the congestion degree of the surrounding area.
In the aspect of traffic flow prediction, the CNN is introduced for spatial modeling, so that the spatial characteristics of traffic flow can be better described. High-dimensional data can be processed more efficiently through the convolution of the CNN, and meanwhile, the spatial characteristics of traffic flow data can be automatically learned, so that the purpose of improving the model prediction accuracy is achieved. Convolutional layers of a convolutional neural network are not fully connected, but are locally connected.
Therefore, the convolution layer can be used for extracting local spatial features, and the same convolution kernel and the input of different positions can be repeatedly calculated by sliding the window, so that the overlarge parameter size is avoided.
The pooling layer selects significant features from the receptive zone, and reduces the sensitivity of the convolutional layer to position while ensuring that the most important features are retained, so the number of model parameters can be greatly reduced compared to artificial neural networks.
CNN is mainly composed of convolution layer and pooling layer, and the formula of convolution operation is defined as:
X l ＝f(W l *C l +b l ) (4)
where f is the activation function, xl and Cl represent the output and input of the first convolutional layer, wl is the weight of the convolutional layer, and bl is the bias of the convolutional layer. The pooling operation formula defines:
C l+1 ＝pooling(X l ) (5)
in the formula, the pooling function represents the pooling operation, with Xl being the input to the pooling layer and Cl +1 being the output of the pooling layer.
The main task of the CNN component is to extract spatial features of urban traffic congestion states based on meshing using convolutional layers in the CNN.
And (3) regarding the urban traffic jam matrix Y under the time slice t as a single-channel image, taking Y as input, performing convolution operation on the Y, and mining local spatial features.
In this embodiment, the two convolutional layers are divided into convolutional layer 1 and convolutional layer 2, wherein:
convolutional layer 1
The convolution operation is carried out on the input 8X8 two-dimensional matrix, local space characteristics can be mined through the convolution operation, the congestion state of the urban area is not only related to the traffic flow of the area of the urban area, but also related to the traffic flow of the surrounding area, the adopted convolution dimension with the dimension of 3-by-3 is set to be 1, the moving step length is set to be 1, and the filling is set to be 1. After the convolution operation, the data is input into a pooling layer to be pooled.
Convolutional layer 2
After the processing of the convolutional layer 1, the urban traffic congestion state matrix is changed from 8x8 to 4x4, then the 4x4 two-dimensional matrix is input into the convolutional layer 2 for convolution operation, the convolution kernel adopts 3 x 3 convolution dimensionality, the moving step length is set to be 1, and the filling is set to be 1. After the convolution operation, the data is input into a pooling layer to be pooled. After the operation of the convolution layer 2, the output matrix size is 2x2, and the input matrix size is input into the position coding assembly after the input matrix size is input into the two full connection layers.
S400, carrying out position coding processing on the obtained spatial characteristic data according to a position coding strategy, wherein the position coding strategy comprises local position coding, global position coding and periodic position coding.
It is understood that step S400 includes steps S401, S402, S403, S404, S405, and S406, where:
s401, sequencing historical road traffic jam state data according to a time sequence to obtain a historical traffic sequence set;
s402, sequencing the road traffic jam state data to be predicted according to a time sequence to obtain a traffic sequence set to be predicted;
s403, calculating a historical traffic sequence set to-be-predicted traffic sequence set based on a relative position strategy to obtain a road traffic sequence set;
s404, inputting the sequence set into a position coding model according to a position coding strategy, calculating to obtain time sequence data, and taking the time sequence data as a local position code;
s405, calculating to obtain a global position code and a periodic position code according to the historical traffic sequence set and the traffic sequence set to be predicted;
and S406, according to the attention mechanism, fusing the local position code, the global position code and the periodic position code to obtain a similarity position combination result.
It should be noted that, since the Transformer model does not include cycles and convolutions, in order for the model to utilize the order of the sequences, some information about the position must be added to the input sequence. Therefore, the position code is added to the input embedding at the bottom of the encoder and decoder stack, and the position coded sequence and the embedded sequence have the same dimension, so that the two can be directly added. The formula for position coding is defined as:
PE (pos,2i) ＝sin(pos/10000 2i/d ) (6)
PE (pos,2i+1) ＝cos(pos/10000 2i/d ) (7)
equation pos and i represent the position index of the input sequence and the dimension position of the input sequence, respectively, and d represents the dimension size of the input sequence. Each dimension of the position code corresponds to a sinusoidal signal, the wavelengths from 2 pi to 10000 · 2 pi being geometrically ordered.
Transformers have had significant success in natural language processing, including machine translation. Although this coding strategy is useful for machine translation, it is not suitable for traffic prediction because temporal characteristics such as continuity and periodicity of time series must be considered for traffic prediction. Applying transformers to traffic prediction requires changing the coding strategy. In machine translation, the input sequence and the target sequence represent sentences of the same meaning in two different languages, and therefore the two sequences should share the same positional index.
However, in the traffic prediction, the history sequence and the prediction sequence are continuous, and thus the history sequence and the prediction sequence have no corresponding relationship. In addition, the traffic data has periodicity, for example, the traffic state of a certain area at three points in the afternoon on the last Monday is similar to the traffic state of the same area at three points in the afternoon on the week. Therefore, when applying the Transformer to traffic prediction, a new strategy is required to encode the time characteristics.
In the embodiment, 15 minutes is adopted as a time slice for the urban area traffic jam state, and the urban traffic jam state { Yn +1, \ 8230;, yn + m } sequence is predicted by assuming that { Y1, Y2 \8230;, yn } is used as the historical urban area traffic jam state sequence.
The method is characterized in that input data is added with other data according to the chronological order, for example, the data is 60 minutes of data, one small data is formed every 15 minutes, the input data consists of 4 small data every 60 minutes, then the four small data have the chronological order, but the transducer does not know that the transducer processes the four data at the same time, the four data do not have the chronological order, and in order to show the chronological order of the 4 data in the transducer, a position coding mode is adopted for achieving the purpose.
It should be noted that, in the present invention, there are three encoding strategies, including local position encoding, global position encoding, and periodic position encoding, where:
1. local position coding strategy: the coding mode is to ensure the local continuity of the time sequence, namely only the continuity between the historical traffic sequence and the predicted traffic sequence is concerned, but not the continuity of the whole traffic sequence. The historical traffic sequence-the predicted traffic sequence is firstly sorted according to time, 0 is used as an initial position, and 1 is added to an index position every time step later. The position indexes are then passed through equations (8) and (9), and the position code of each time series is obtained.
For example, { Y1, Y2 \8230;, yn } is the historical traffic sequence, { Yn +1, \8230;, yn + m } is the traffic sequence to be predicted, and the index positions obtained using the relative position strategy are (0, 1, \8230;, n-1) and (n, \8230;, n + m-1), respectively.
2. Global position coding strategy: the local position strategy can ensure the local continuity of the sequence, but also neglects that a time segment sequence is in different historical traffic sequences-the predicted traffic sequence is actually the same time segment sequence. For example, the traffic congestion status sequence with time at 3. Therefore, the local position strategy is limited in that the same time series cannot be made to show the same position code in different historical traffic sequences, predicted traffic sequences. Therefore, the global position strategy is provided to solve the limitation of the local position strategy and ensure the global continuity of the time sequence.
The global position strategy can ensure that only one position code exists in the time slices in the whole time period even if the time slices occur in different sequences, and uniqueness is ensured. First, all time slices in the data set are sorted according to time, then the indexes are carried out from 0,1 is added to each time slice index, and then all the indexes are substituted into a formula (121) to obtain the global position code of each time slice.
3. Periodic position encoding strategy: in the traffic sequence, not only the time characteristics of continuity but also the time characteristics of periodicity, including day periodicity and week periodicity, need to be considered. By adding the periodic time characteristics, the model can be ensured to have better prediction accuracy. The periodicity comprises a day period and a week period, and the day periodicity can be ensured by adding time slices of the same time period of the previous day and the predicted traffic sequence into the historical traffic sequence and sequencing along a time axis. The week periodicity is ensured by adding time slices of the same time period of the same week of the previous week and the predicted traffic sequence into the historical traffic sequence and then sequencing along the time axis. Let { Y1, Y2 \8230;, yn } be the historical traffic sequence, and { Yn +1, \8230;, yn + m } be the traffic sequence to be predicted.
For day periodicity, time slice data { D1, \8230;, dm } of the previous day and the same time period of the predicted traffic sequence is sought. For the week periodicity, time slice data { W1, \8230;, wm } for the same time period of the same week of the previous week is sought. The day-periodic time period and the week-periodic time period are spliced with the historical traffic sequence according to the time axis to serve as a new historical traffic sequence, namely { W1, \8230;, wm, D1, \8230;, dm, Y1, Y2, \8230;, yn }, and then { Yn +1, \8230;, yn + m } is predicted.
The historical traffic sequence-the predicted traffic sequence are respectively subjected to local position coding strategy and global position coding strategy to obtain two position codes, the two position codes have the same dimension and can be directly added, the added result is used as a new position code pos, and the position codes not only consider the locality of the sequence, but also consider the global nature of the sequence.
Wherein the similarity position combination is a mode of directly adding the position code pos to the historical traffic sequence-predicted traffic sequence input into the model, and adding some information about the position to the input sequence. However, since the position code pos includes both local position code and global position code, if some information about position is added to the input sequence by directly adding to the input sequence, other relationships between the sequences may be hidden, which makes training the model more difficult, and reduces the accuracy of prediction, especially when the data set is small.
Therefore, an improved mode is provided, the position codes are better fused into the input sequence, and not only some information about the position is added into the input sequence, but also the position codes can be ensured not to hide other relations among the sequences, so that the model is more difficult to train. And a position combination mode based on an attention mechanism is added to the CNN space component and the Transformer time component, and the prediction accuracy is improved by further improving the CNN-Transformer model. The calculation formula is as follows:
S＝softmax(pos×pos T ) (8)
the equation pos is the position code of the input sequence, Q, K and V are respectively the query vector, the key vector and the value vector after the input sequence is subjected to linear transformation, and dk is the dimension of K.
And S500, inputting the processed position encoding data into a Transformer model for prediction based on an attention mechanism to obtain a road traffic jam prediction result.
It is understood that step S500 includes steps S501, S502, S503, S505, and S506, where:
s501, inputting the combined result of the similarity positions into a Transformer model, and calculating correlation values among different sequences according to a dot product function;
s502, according to the correlation value, performing logistic regression function calculation on each sequence to obtain a logistic regression value of each sequence;
s503, inputting the logistic regression value of each sequence into a residual error neural network, and adding the logistic regression value of each sequence and the corresponding sequence data to obtain fusion data;
s504, inputting the fusion data into a feedforward neural network for calculation according to the feedforward neural network to obtain intermediate data;
s505, inputting the obtained intermediate data to a residual neural network again, and adding the intermediate data and the fusion data to obtain result data;
and S506, predicting the result data according to the encoder of the Transformer, and calculating to obtain a road traffic jam prediction result.
It should be noted that the Transformer is developed as a new deep learning framework, and performs sequence modeling based on attention mechanism and position coding strategy. The Transformer builds entirely on the attention mechanism, which enables it to visit any part of the sequence, regardless of its distance from the target. Essentially, the Transformer is constructed in the manner of an encoding component, which is partly constituted by a stack of encoders, a decoding component, which is constituted by a stack of decoders, the number of encoders and decoders being the same. Each encoder module is composed of a multi-head self-attention layer and a position feedforward layer, and each decoder module is provided with an encoder-decoder attention layer which is inserted between the self-attention layer and the feedforward layer and used as a bridge of the encoder and the decoder.
Attention mechanism is a research that has been generated due to the emergence of information processing bottlenecks. It can selectively focus on a portion of the information, ignore other visible information, and decide which portion of the input to focus on, thereby allocating limited resources to the important portion. The attention function is the mapping of a query and a set of key-value pairs to an output. The output of the function is a weighted sum, and the weight assigned to each value is obtained by calculation of the compatible function of the query and the corresponding key.
The multi-headed attention mechanism is a variant of the attention mechanism that can compute multiple pieces of information from an input sequence in parallel. Each focusing on a different part of the input sequence in order to notice information in a different subspace, capturing more abundant characteristic information. The multi-point attention mechanism calculates the correlation between input sequences to focus on more important information. For example, a multi-head attention mechanism operates on the inputs (a 1, a2, \ 8230;, an), which calculates the correlation of each input sequence with other input sequences, and then outputs the same number of sequences (y 1, y2, \8230;, yn).
The calculation steps of the multi-head attention mechanism are as follows:
(1) Firstly, calculating correlation values among different sequences by using a dot product function, wherein the formula is as follows:
in the formula, ai and aj represent sequences input into a multi-head attention mechanism; WQ and WK are linear transformation matrixes; d is the dimension of aj WK.
(2) A logistic regression model was performed on the correlation values α ij of each sequence with the other sequences.
(3) The output of the multi-head attention mechanism is calculated, and the bi dimension is the same as the input value ai.
In the formula, p ij For the values calculated by equation (11), wv is a linear transformation matrix, and aj is the sequence of the input multi-headed attention mechanism.
In addition to the attention layer, a feed-forward neural network is included at each encoder and decoder. Two linear transformations are included in the feedforward neural network, one of which is the ReLU activation function.
FFN(x)＝max(0,xW 1 +b 1 )W 2 +b 2 (13)
In the formula, x is a value input into the feedforward neural network, max () refers to a maximum value of two parameters, W1, W2, b1 and b2 are all weight values, the input and output dimensions of the feedforward neural network are 512, and the inner dimension is 2048.
Example 2:
as shown in fig. 2, the present embodiment provides a road traffic congestion prediction apparatus, referring to fig. 2, the apparatus includes an acquisition module 701, a calculation module 702, an extraction module 703, a processing module 704, and a prediction module 705, where:
the collecting module 701 is configured to collect first data, where the first data is road condition data of each road in a road traffic network in different time periods.
Preferably, the acquisition module 701 includes a dividing unit 7011, a first obtaining unit 7012, a matching unit 7013, a first calculating unit 7014, and a second calculating unit 7015, wherein:
a dividing unit 7011, configured to divide road traffic grid regions of the city to obtain road network data of the city;
the first obtaining unit 7012 is configured to obtain first positioning data, where the first positioning data is original positioning data generated by a driver in a process of using the mobile phone navigation software, where the original positioning data includes field information of time, longitude and latitude, speed, and direction angle;
a matching unit 7013, configured to match the first positioning data to road network data of the city to obtain a positioning track after matching the road network;
the first calculating unit 7014 is configured to calculate a traffic state of each road in a fixed time period based on the positioning track;
and the second calculating unit 7015 is configured to calculate, according to the traffic state, a traffic state of each divided area in the road traffic grid of the city.
And the calculating module 702 is used for calculating matrix data of the traffic jam indexes of each road in different time periods according to the first data.
Preferably, the calculation module 702 includes a second obtaining unit 7021, a third calculation unit 7022, a constructing unit 7023, a fourth calculation unit 7024, a solving unit 7025, and a fifth calculation unit 7026, wherein:
a second obtaining unit 7021, configured to obtain a traffic flow and an average traveling speed of a vehicle for each road segment in the road network according to the data provided by the road network vehicle detection device;
a third calculating unit 7022, configured to calculate, according to the traffic flow of each road segment in the road network and the average traveling speed of the vehicle, the number of vehicles included in each road segment in the road network;
a constructing unit 7023, configured to construct a congestion intensity function with speed as an independent variable;
a fourth calculating unit 7024, configured to input the speed of each road segment to the congestion intensity function, and calculate an average congestion intensity of the road segment;
the solving unit 7025 is configured to use the number of vehicles in each road segment as a weight of the road segment, and according to the average congestion intensity, solve a weighted average of the congestion intensity of the whole road network;
and a fifth calculating unit 7026, configured to use the weighted average as a traffic congestion index of the road network, collect road segment length information, and calculate to obtain matrix data of the traffic congestion index.
The extraction module 703 is configured to input at least one matrix data into a preset prediction model based on a convolutional neural network algorithm, and extract at least one spatial feature data, where the spatial feature data includes a congestion factor for each road.
Preferably, the extracting module 703 includes a setting unit 7031, a first convolution unit 7032, a mining unit 7033, a second convolution unit 7034, a constructing unit 7035, and an extracting unit 7036, where:
a setting unit 7031 for setting the matrix data to a two-dimensional matrix of 8X 8;
a first convolution unit 7032, configured to perform a first convolution on the 8X8 two-dimensional matrix based on a convolutional neural network algorithm;
the mining unit 7033 is configured to mine the two-dimensional matrix after the first convolution operation to obtain a 4X4 two-dimensional matrix;
a second convolution unit 7034, configured to perform a second convolution on the 4X4 two-dimensional matrix to obtain a 2X2 two-dimensional matrix;
a constructing unit 7035 configured to construct a position coding model;
and the extracting unit 7036 is configured to, based on a convolutional neural network algorithm, input the 2X2 two-dimensional matrix into the position coding model after passing through the fully-connected neural network, and extract spatial feature data of at least one current traffic congestion vicinity area.
And the processing module 704 is configured to perform position coding processing on the obtained spatial feature data according to a position coding strategy, where the position coding strategy includes local position coding, global position coding, and periodic position coding.
Preferably, processing module 704 comprises a first sorting unit 7041, a second sorting unit 7042, a sixth calculating unit 7043, a seventh calculating unit 7044, an eighth calculating unit 7045, and a fusing unit 7046, wherein:
first sorting unit 7041: the system comprises a historical road traffic jam state data acquisition unit, a historical traffic sequence set and a traffic data processing unit, wherein the historical road traffic jam state data acquisition unit is used for sequencing historical road traffic jam state data according to a time sequence to obtain a historical traffic sequence set;
second sorting unit 7042: the system is used for sequencing the road traffic jam state data to be predicted according to a time sequence to obtain a traffic sequence set to be predicted;
sixth calculating unit 7043: the system comprises a traffic sequence set to be predicted, a historical traffic sequence set and a traffic sequence set to be predicted, wherein the traffic sequence set to be predicted is used for calculating based on a relative position strategy to obtain a road traffic sequence set;
seventh calculating unit 7044: the system comprises a position coding model, a position coding strategy and a position coding strategy, wherein the position coding model is used for inputting a sequence set into the position coding model according to the position coding strategy, calculating to obtain time sequence data, and taking the time sequence data as a local position code;
eighth calculating unit 7045: the system comprises a traffic sequence set to be predicted, a global position code and a periodic position code, wherein the traffic sequence set to be predicted is used for calculating according to a historical traffic sequence set;
fusion unit 7046: and the method is used for fusing the local position code, the global position code and the periodic position code according to an attention mechanism to obtain a similarity position combination result.
And the prediction module 705 is configured to input the processed position encoding data into a Transformer model for prediction based on an attention mechanism, so as to obtain a road traffic congestion prediction result.
Preferably, the prediction module 705 comprises a ninth calculation unit 7051, a tenth calculation unit 7052, a first input unit 7053, a second input unit 7054, a third input unit 7055, and a prediction unit 7056, wherein:
ninth calculating unit 7051: the method is used for inputting the combined result of the similarity positions into a Transformer model and calculating correlation values among different sequences according to a dot product function;
tenth calculating unit 7052: the correlation value calculating module is used for calculating a logistic regression function of each sequence according to the correlation value to obtain a logistic regression value of each sequence;
first input unit 7053: the system is used for inputting the logistic regression value of each sequence into a residual error neural network, and adding the logistic regression value of each sequence and the corresponding sequence data thereof to obtain fusion data;
second input unit 7054: the fusion data are input to the feedforward neural network for calculation according to the feedforward neural network to obtain intermediate data;
third input unit 7055: the residual error neural network is used for inputting the obtained intermediate data into the residual error neural network again, and adding the intermediate data and the fusion data to obtain result data;
prediction unit 7056: and the method is used for predicting the result data according to the encoder of the Transformer and calculating to obtain a road traffic jam prediction result.
It should be noted that, regarding the apparatus in the above embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment related to the method, and will not be elaborated herein.
Example 3:
corresponding to the above method embodiments, the present embodiment further provides a road traffic jam prediction device, and a road traffic jam prediction device described below and a road traffic jam prediction method described above may be referred to in correspondence with each other.
Fig. 3 is a block diagram illustrating a road traffic congestion prediction apparatus 800, according to an example embodiment. As shown in fig. 3, the road traffic congestion prediction apparatus 800 may include: a processor 801, a memory 802. The road traffic congestion prediction apparatus 800 may further include one or more of a multimedia component 803, an i/O interface 805, and a communication component 805.
The processor 801 is configured to control the overall operation of the road traffic congestion prediction apparatus 800, so as to complete all or part of the steps of the road traffic congestion prediction method. The memory 802 is used to store various types of data to support operation of the road traffic congestion prediction device 800, such as instructions for any application or method operating on the road traffic congestion prediction device 800, and application-related data, such as contact data, transceived messages, pictures, audio, video, and so forth. The Memory 802 may be implemented by any type of volatile or non-volatile Memory device or combination thereof, such as Static Random Access Memory (SRAM), electrically Erasable Programmable Read-Only Memory (EEPROM), erasable Programmable Read-Only Memory (EPROM), programmable Read-Only Memory (PROM), read-Only Memory (ROM), magnetic Memory, flash Memory, magnetic disk or optical disk. The multimedia components 803 may include screen and audio components. Wherein the screen may be, for example, a touch screen and the audio component is used for outputting and/or inputting audio signals. For example, the audio component may include a microphone for receiving an external audio signal. The received audio signal may further be stored in the memory 802 or transmitted through the communication component 805. The audio assembly also includes at least one speaker for outputting audio signals. The I/O interface 805 provides an interface between the processor 801 and other interface modules, such as a keyboard, mouse, buttons, and the like. These buttons may be virtual buttons or physical buttons. The communication component 805 is used for wired or wireless communication between the road traffic congestion prediction device 800 and other devices. Wireless communication, such as Wi-Fi, bluetooth, near field communication (NFC for short), 2G, 3G, or 5G, or a combination of one or more of them, so the corresponding communication component 805 may include: wi-Fi module, bluetooth module, NFC module.
In an exemplary embodiment, the road traffic congestion prediction apparatus 800 may be implemented by one or more Application Specific Integrated Circuits (ASICs), digital Signal Processors (DSPs), digital Signal Processing Devices (DSPDs), programmable Logic Devices (PLDs), field Programmable Gate Arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components for performing the above-described road traffic congestion prediction methods.
In another exemplary embodiment, a computer-readable storage medium is also provided, which comprises program instructions, which when executed by a processor, implement the steps of the above-described road traffic congestion prediction method. For example, the computer readable storage medium may be the above-described memory 802 including program instructions executable by the processor 801 of the road traffic congestion prediction apparatus 800 to perform the above-described road traffic congestion prediction method.
Example 5:
corresponding to the above method embodiments, a readable storage medium is also provided in this embodiment, and a readable storage medium described below and a road traffic congestion prediction method described above may be referred to in correspondence with each other.
A readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method for road traffic congestion prediction of the above-mentioned method embodiments.
The readable storage medium may be a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and various other readable storage media capable of storing program codes.
In conclusion, the invention provides the prediction of applying CNN in combination with a Transformer to urban traffic, so that the training speed of the model can be greatly improved, the hidden danger of gradient problem of CNN-LSTM is overcome, and the accuracy of the model is further improved; a position coding component is added to the CNN-Transformer model. The manner in which the position-coded pos is directly added to the historical-predicted traffic sequence input into the Transformer component has certain drawbacks in adding some information about the position to the input sequence. Since the position code pos includes both local position coding and global position coding, if some information about position is added to the input sequence by directly adding to the input sequence, other relations between the sequences may be hidden, which makes training the model more difficult and reduces the prediction accuracy, especially when the data set is small.
Therefore, the invention provides an improved mode, namely a position coding component is added, the position coding component can be better fused into the input sequence, and not only some information about the position is added into the input sequence, but also the position coding component can ensure that other relations among the sequences are not hidden, so that the model training is more difficult. And a position combination mode based on an attention mechanism is added to the CNN space component and the Transformer time component, and the accuracy of prediction is improved by further improving the CNN-Transformer model.
The above is only a preferred embodiment of the present invention, and is not intended to limit the present invention, and various modifications and changes will occur to those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention, and the changes or substitutions should be covered within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A method for predicting road traffic congestion, comprising:
acquiring first data, wherein the first data are road condition data of each road in a road traffic network in different time periods;
calculating matrix data of traffic jam indexes of each road in different time periods according to the first data;
inputting at least one matrix data into a preset prediction model based on a convolutional neural network algorithm, and extracting at least one spatial feature data, wherein the spatial feature data comprises congestion factors of each road;
according to a position coding strategy, carrying out position coding processing on the obtained spatial feature data, wherein the position coding strategy comprises local position coding, global position coding and periodic position coding;
and inputting the processed position encoding data into a Transformer model for prediction based on an attention mechanism to obtain a road traffic jam prediction result.
2. The method according to claim 1, wherein the collecting first data is road condition data of each road in a road traffic network in different time periods, and comprises:
dividing a road traffic grid region of a city, and acquiring road network data of the city;
acquiring first positioning data, wherein the first positioning data is original positioning data generated by a driver in a process of using mobile phone navigation software, and the original positioning data comprises field information of time, longitude and latitude, speed and direction angle;
matching the first positioning data to the road network data of the city to obtain a positioning track matched with a road network;
calculating the traffic state of each road in a fixed time period based on the positioning track;
and calculating the traffic state of each divided region in the road traffic grid of the city according to the traffic state.
3. The method for predicting road traffic congestion according to claim 1, wherein the calculating matrix data of the traffic congestion indicators of each road in different time periods according to the first data comprises:
obtaining the traffic flow and the average running speed of vehicles of each road section in the road network according to the data provided by the road network vehicle detection equipment;
calculating the number of vehicles contained in each road section in the road network according to the traffic flow of each road section in the road network and the average driving speed of the vehicles;
constructing a congestion intensity function with speed as an independent variable;
inputting the speed of each road section into the congestion intensity function, and calculating to obtain the average congestion intensity of the road section;
the number of vehicles of each road section is used as the weight of the road section, and the weighted average value of the congestion intensity of the whole road network is solved according to the average congestion intensity;
and taking the weighted average value as a traffic congestion index of the road network, collecting road section length information, and calculating to obtain the matrix data of the traffic congestion index.
4. The method for predicting road traffic congestion according to claim 1, wherein the step of inputting at least one matrix data into a preset prediction model based on a convolutional neural network algorithm to extract at least one spatial feature data comprises:
setting the matrix data to be a two-dimensional matrix of 8X 8;
performing a first convolution on the 8X8 two-dimensional matrix based on the convolutional neural network algorithm;
digging the two-dimensional matrix after the first convolution operation to obtain a 4X4 two-dimensional matrix;
performing second convolution on the two-dimensional matrix of 4X4 to obtain a two-dimensional matrix of 2X 2;
constructing a position coding model;
based on the convolutional neural network algorithm, the 2X2 two-dimensional matrix is input into the position coding model after passing through a full-connection neural network, and spatial feature data of at least one current traffic jam nearby area are extracted.
5. A road traffic congestion prediction apparatus, comprising:
the acquisition module is used for acquiring first data, wherein the first data are road condition data of each road in a road traffic network in different time periods;
the calculation module is used for calculating matrix data of traffic jam indexes of each road in different time periods according to the first data;
the extraction module is used for inputting at least one matrix data into a preset prediction model based on a convolutional neural network algorithm and extracting at least one spatial characteristic data, wherein the spatial characteristic data comprises congestion factors of each road;
the processing module is used for carrying out position coding processing on the obtained spatial characteristic data according to a position coding strategy, wherein the position coding strategy comprises local position coding, global position coding and periodic position coding;
and the prediction module is used for inputting the processed position encoding data into a Transformer model for prediction based on an attention mechanism to obtain a road traffic jam prediction result.
6. The apparatus for predicting road traffic congestion according to claim 5, wherein the collecting module comprises:
the dividing unit is used for dividing road traffic grid areas of a city and acquiring road network data of the city;
the first acquisition unit is used for acquiring first positioning data, wherein the first positioning data is original positioning data generated by a driver in the process of using mobile phone navigation software, and the original positioning data comprises field information of time, longitude and latitude, speed and direction angle;
the matching unit is used for matching the first positioning data to the road network data of the city to obtain a positioning track matched with a road network;
the first calculation unit is used for calculating and obtaining the traffic state of each road in a fixed time period based on the positioning track;
and the second calculation unit is used for calculating the traffic state of each divided area in the road traffic grid of the city according to the traffic state.
7. The apparatus for predicting road traffic congestion according to claim 5, wherein the calculating means comprises:
the second acquisition unit is used for acquiring the traffic flow and the average driving speed of the vehicle of each road section in the road network according to the data provided by the road network vehicle detection equipment;
the third calculating unit is used for calculating the number of vehicles contained in each road section in the road network according to the traffic flow of each road section in the road network and the average running speed of the vehicles;
the construction unit is used for constructing a congestion intensity function with speed as an independent variable;
the fourth calculation unit is used for inputting the speed of each road section into the congestion intensity function and calculating the average congestion intensity of the road section;
the solving unit is used for solving the weighted average value of the congestion intensity of the whole road network according to the average congestion intensity by using the number of vehicles of each road section as the weight of the road section;
and the fifth calculation unit is used for taking the weighted average value as a traffic congestion index of the road network, acquiring road section length information and calculating to obtain the matrix data of the traffic congestion index.
8. The apparatus for predicting road traffic congestion according to claim 5, wherein the extracting module comprises:
the setting unit is used for setting the matrix data into a two-dimensional matrix of 8X 8;
the first convolution unit is used for performing first convolution on the 8X8 two-dimensional matrix based on the convolution neural network algorithm;
the digging unit is used for digging the two-dimensional matrix after the first convolution operation to obtain a 4X4 two-dimensional matrix;
the second convolution unit is used for carrying out second convolution on the 4X4 two-dimensional matrix to obtain a 2X2 two-dimensional matrix;
the construction unit is used for constructing a position coding model;
and the extraction unit is used for inputting the 2X2 two-dimensional matrix into the position coding model after passing through the fully-connected neural network based on the convolutional neural network algorithm, and extracting spatial characteristic data of at least one current traffic jam nearby area.
9. A road traffic congestion prediction apparatus characterized by comprising:
a memory for storing a computer program;
a processor for implementing the steps of the road traffic congestion prediction method according to any one of claims 1 to 4 when executing said computer program.
10. A readable storage medium, characterized by: the readable storage medium has stored thereon a computer program which, when being executed by a processor, carries out the steps of the method for predicting road traffic congestion according to any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210604273.1A CN115222089A (en) | 2022-05-30 | 2022-05-30 | Road traffic jam prediction method, device, equipment and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210604273.1A CN115222089A (en) | 2022-05-30 | 2022-05-30 | Road traffic jam prediction method, device, equipment and readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115222089A true CN115222089A (en) | 2022-10-21 |
Family
ID=83607078
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210604273.1A Pending CN115222089A (en) | 2022-05-30 | 2022-05-30 | Road traffic jam prediction method, device, equipment and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115222089A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115620524A (en) * | 2022-12-15 | 2023-01-17 | 中南大学 | Traffic jam prediction method, system, equipment and storage medium |
| CN117691592A (en) * | 2023-12-14 | 2024-03-12 | 湖南防灾科技有限公司 | Photovoltaic output prediction method and device, electronic equipment and storage medium |
-
2022
- 2022-05-30 CN CN202210604273.1A patent/CN115222089A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115620524A (en) * | 2022-12-15 | 2023-01-17 | 中南大学 | Traffic jam prediction method, system, equipment and storage medium |
| CN117691592A (en) * | 2023-12-14 | 2024-03-12 | 湖南防灾科技有限公司 | Photovoltaic output prediction method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Xu et al. | Real-time prediction of taxi demand using recurrent neural networks | |
| CN112470199A (en) | System and method for point-to-point traffic prediction | |
| CN115222089A (en) | Road traffic jam prediction method, device, equipment and readable storage medium | |
| Feng | Modeling dynamic urban land-use change with geographical cellular automata and generalized pattern search-optimized rules | |
| CN113204718A (en) | Vehicle track destination prediction method considering space-time semantics and driving state | |
| CN113989330A (en) | Vehicle track prediction method and device, electronic equipment and readable storage medium | |
| CN113570867B (en) | Urban traffic state prediction method, device, equipment and readable storage medium | |
| Liang et al. | NetTraj: A network-based vehicle trajectory prediction model with directional representation and spatiotemporal attention mechanisms | |
| CN113159403A (en) | Method and device for predicting pedestrian track at intersection | |
| CN112559585A (en) | Traffic space-time sequence single-step prediction method, system and storage medium | |
| CN114202120A (en) | Urban traffic travel time prediction method aiming at multi-source heterogeneous data | |
| CN112990530B (en) | Regional population quantity prediction method, regional population quantity prediction device, electronic equipment and storage medium | |
| Zhou et al. | An attention-based deep learning model for citywide traffic flow forecasting | |
| Qian et al. | Vehicle trajectory modelling with consideration of distant neighbouring dependencies for destination prediction | |
| CN117610734A (en) | Deep learning-based user behavior prediction method, system and electronic equipment | |
| Guan et al. | A trajectory prediction method based on bayonet importance encoding and bidirectional LSTM | |
| Al-Yarimi | Enhancing road safety through advanced predictive analytics in V2X communication networks | |
| Yao et al. | Goal-LBP: Goal-Based Local Behavior Guided Trajectory Prediction for Autonomous Driving | |
| CN115146844A (en) | Multi-mode traffic short-time passenger flow collaborative prediction method based on multi-task learning | |
| Osipov et al. | Neural network forecasting of traffic congestion | |
| CN114154711A (en) | Traffic information recommendation method and device, electronic equipment and computer storage medium | |
| Fatkhulin et al. | Analysis of the Basic Image Generation Methods by Neural Networks | |
| CN113706347A (en) | Multitask model distillation method, multitask model distillation system, multitask model distillation medium and electronic terminal | |
| CN112116167A (en) | Data processing method, data processing apparatus, storage medium, and electronic device | |
| CN116562487B (en) | Moving destination prediction method considering intersection space-time correlation and history travel semantics |
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 |