A traffic indicator sensitivity analysis method, device, equipment and medium

Abstract

The application is suitable for the field of intelligent transportation technology, and provides a traffic index sensitivity analysis method, device, equipment and medium. The traffic index sensitivity analysis method comprises: acquiring time index data, space index data and road network index data; constructing a traffic state evaluation model; using a time feature extraction module to extract features from the time index data to obtain time features, using a space feature extraction module to extract features from the space index data to obtain space features, and using a space-time feature fusion module to fuse the features to obtain space-time fusion features; using a road network feature extraction module to extract features from the space-time fusion features and the road network index data to obtain final traffic features; using an evaluation module to evaluate the final traffic features to obtain a traffic state; and calculating the sensitivity of the time index, the space index and the road network index based on the traffic state. The traffic index sensitivity analysis method can improve the rationality of the sensitivity.

Description

Technical Field

[0001] This application relates to the field of intelligent transportation technology, and in particular to a method, apparatus, equipment and medium for analyzing the sensitivity of traffic indicators. Background Technology

[0002] During traffic operations, the traffic status of a regional road network is influenced by numerous traffic parameters, such as traditional parameters like total road length, total area, road network geometry, proportion and spatial layout of roads of different grades, length and layout of bus lanes, and intersection signal control schemes. Traffic status itself can also be reflected by several parameters, such as road network capacity, saturation, vehicle speed, travel time, and congestion level. Assessing regional traffic status is crucial for user travel planning, business site selection, and traffic management. With increasing emphasis on the intelligent connected vehicle industry and vehicle-road cooperative systems, the deployment density of related intelligent connected equipment is gradually increasing. Through vehicle-road intelligent connected systems, complete vehicle operation data can be collected, providing richer and more accurate data for regional traffic status assessment.

[0003] Intelligent connected traffic data is traffic data acquired through intelligent connected roadside sensors, including vehicle information, traffic flow information, and collection time. It can effectively record the spatiotemporal evolution of traffic patterns within the road network. Intelligent connected traffic data has advantages such as high accuracy, comprehensiveness, and large volume. However, existing research on models and methods based on intelligent connected traffic data often focuses on data security, lane changing, and traffic flow prediction, with limited research on regional traffic state assessment. Therefore, this invention proposes a regional traffic state assessment technology based on intelligent connected traffic data, and conducts sensitivity analysis on this basis. This technology can effectively utilize intelligent connected traffic data to identify key factors affecting regional traffic conditions, providing a scientific basis for traffic managers to formulate management strategies.

[0004] Overall, existing research has explored regional traffic condition assessment to some extent, primarily employing methods such as the analytic hierarchy process (AHP), fuzzy comprehensive evaluation, and neural network methods. However, the evaluation indicators considered in these studies are relatively macroscopic, lacking attention to some important micro-level indicators. They also fail to fully utilize intelligent connected traffic data, making it difficult to construct a scientifically sound and reasonable evaluation indicator system. Furthermore, after obtaining the assessment results, existing studies often only conduct qualitative analysis, lacking further quantitative and precise calculations, thus lacking persuasiveness. Therefore, there is a problem of unreasonable analysis of traffic indicators. Summary of the Invention

[0005] This application provides a method, apparatus, device, and medium for traffic indicator sensitivity analysis, which can solve the problem of unreasonable analysis of traffic indicators.

[0006] In a first aspect, embodiments of this application provide a traffic indicator sensitivity analysis method, which includes:

[0007] Acquire time index data of multiple time indicators, spatial index data of multiple spatial indicators, and road network index data of multiple road network indicators for the target area at time T; the Tth time is the current time.

[0008] A traffic condition assessment model is constructed. The traffic condition assessment model includes: a time feature extraction module, a spatial feature extraction module, and a sequentially connected time-spatial feature fusion module, a road network feature extraction module, and an assessment module.

[0009] The time feature extraction module is used to extract features from all time indicator data to obtain time features. The spatial feature extraction module is used to extract features from all spatial indicator data to obtain spatial features. Finally, the spatiotemporal feature fusion module is used to fuse the time features and spatial features to obtain spatiotemporal fused features.

[0010] The road network feature extraction module is used to extract features from the spatiotemporal fusion features and all road network indicator data of the target area to obtain the final traffic features of the target area.

[0011] The final traffic characteristics are evaluated using the evaluation module to obtain the traffic status of the target area at the current moment;

[0012] Based on the traffic status of the target area at the current time, calculate the time sensitivity of each time indicator, the spatial sensitivity of each spatial indicator, and the road network sensitivity of each road network indicator at the current time.

[0013] Optional, the time feature extraction module includes:

[0014] A fully convolutional neural network unit and a temporal attention unit connected in sequence;

[0015] The input of the fully convolutional neural network unit is the input of the temporal feature extraction module, and the output of the temporal attention unit is the output of the temporal feature extraction module.

[0016] The spatial feature extraction module includes: a spatial embedding unit and a spatial attention unit connected in sequence;

[0017] The input of the spatial embedding unit is the input of the spatial feature extraction module, and the output of the spatial attention unit is the output of the spatial feature extraction module.

[0018] Optionally, the spatiotemporal feature fusion module includes:

[0019] A global attention unit and multiple sequentially connected spatiotemporally gated recursive units;

[0020] The input of each spatiotemporal gated recursive unit is connected to the output of the temporal feature extraction module and the output of the spatial feature extraction module. The output of each spatiotemporal gated recursive unit is connected to the input of the global attention unit. The output of the global attention unit is the output of the spatiotemporal feature fusion module.

[0021] The road network feature extraction module includes: a road network embedding unit and a gated residual attention unit;

[0022] The input of the road network embedding unit is connected to all road network index data. The input of the gated residual attention unit is connected to the output of the spatiotemporal feature fusion module and the output of the road network embedding unit. The output of the gated residual attention unit is the output of the road network feature extraction module.

[0023] The assessment module includes: a traffic condition assessment unit;

[0024] The input terminal of the traffic condition assessment unit is the input terminal of the assessment module, and the output terminal of the traffic condition assessment unit is the output terminal of the assessment module.

[0025] Optionally, a spatiotemporal feature fusion module can be used to fuse temporal and spatial features to obtain spatiotemporal fused features, including:

[0026] Based on the hidden fusion feature H corresponding to the (f-1)th spatiotemporal gated recursive unit f-1 Two different updates and resets are performed on the time features to obtain the first and second time hidden features corresponding to the f-th spatiotemporal gated recursive unit, respectively.

[0027] Based on the hidden fusion feature H corresponding to the (f-1)th spatiotemporal gated recursive unit f-1 Two different updates and resets are performed on the spatial features to obtain the first spatial hidden feature and the second spatial hidden feature corresponding to the f-th spatiotemporal gated recursive unit, respectively.

[0028] Initial fusion of temporal and spatial features yields initial fused features.

[0029] Through the formula:

[0030]

[0031]

[0032] Calculate the integrated time features corresponding to the f-th spatiotemporally gated recursive unit.

[0033] Among them, w zThe time feature weights are represented by sigmoid(), which represents the activation function. This represents the first-time hidden feature corresponding to the f-th spatiotemporally gated recursive unit. Let f represent the second temporal hidden feature corresponding to the f-th spatiotemporally gated recursive unit, where f = 1, 2, ..., F, and F represents the total number of spatiotemporally gated recursive units. When f = 1, the hidden fusion feature H corresponding to the (f-1)-th spatiotemporally gated recursive unit is... f-1 The value is 0. This represents the weight matrix corresponding to the hidden features in the first space. This represents the weight matrix corresponding to the hidden features in the second space. The symbol ⊙ indicates the bias, and ⊙ indicates the Hadamah product operation.

[0034] Through the formula:

[0035]

[0036]

[0037] Calculate the integrated spatial features corresponding to the f-th spatiotemporally gated recursive unit.

[0038] Among them, w r Represents spatial feature weights, This represents the first spatial hidden feature corresponding to the f-th spatiotemporally gated recursive unit. This represents the second spatial hidden feature corresponding to the f-th spatiotemporally gated recursive unit. This represents the weight matrix corresponding to the hidden features in the first space. This represents the weight matrix corresponding to the hidden features in the second space. Indicates bias;

[0039] Through the formula:

[0040]

[0041]

[0042] Calculate the hidden fusion feature H corresponding to the f-th spatiotemporally gated recursive unit. f ;

[0043] in, This represents the initial hidden fusion feature corresponding to the f-th spatiotemporally gated recursive unit. Indicates the initial fusion features, This represents the weight matrix corresponding to the initial fused features. This represents the weight matrix of the hidden fusion features corresponding to the (f-1)th spatiotemporally gated recursive unit. Indicates bias;

[0044] Obtain the hidden fusion features H = {H1, H2, ..., H} corresponding to all spatiotemporal gated recursive units. F};

[0045] Where H1 represents the hidden fusion feature corresponding to the first spatiotemporally gated recursive unit, H2 represents the hidden fusion feature corresponding to the second spatiotemporally gated recursive unit, and H... F This represents the hidden fusion feature corresponding to the Fth spatiotemporal gated recursive unit;

[0046] Based on the hidden fusion features H corresponding to all spatiotemporal gated recursive units, spatiotemporal fusion features are obtained using global attention units.

[0047] Optionally, the road network feature extraction module is used to extract features from the spatiotemporal fusion features and all road network indicator data of the target area to obtain the final traffic features of the target area, including:

[0048] Feature extraction is performed on all road network indicator data to obtain road network features;

[0049] Through the formula:

[0050] η1=ELU(H * W 1,1 +C'W 1,2 +b1)

[0051] η2=η1W2+b2

[0052]

[0053] Calculate the final traffic characteristics

[0054] Where η1 and η2 both represent intermediate quantities, H * W represents the spatiotemporal fusion characteristics. 1,1 and W 1,2 All represent weight matrices, C' represents road network features, b1 and b2 both represent biases, LN() represents normalization operation, GAU() represents gated layer operation, and ELU() represents exponential linear unit activation function.

[0055] Optionally, the final traffic characteristics are evaluated using the evaluation module in the traffic state assessment model to obtain the traffic state of the target area at the current time, including:

[0056] The evaluation module is used to calculate the final traffic characteristics, and to obtain multiple candidate traffic states for the target area and the probability value corresponding to each candidate traffic state.

[0057] The probability value with the largest value among all probability values ​​is selected as the target probability value, and the candidate traffic state corresponding to the target probability value is taken as the traffic state of the target area.

[0058] Optionally, based on the traffic state of the target area at the current time, calculate the temporal sensitivity of each time indicator, the spatial sensitivity of each spatial indicator, and the road network sensitivity of each road network indicator at the current time, including:

[0059] Traffic status is as follows:

[0060]

[0061] in, C represents the traffic state value, and C represents the data of all traffic indicators in the target area at time T. Represents a constant term. This represents the mapping relationship between the i-th traffic indicator and the traffic state value output by the traffic state assessment model, where i, h ∈ {1, 2, ..., n}, and n represents the total number of traffic indicators. A traffic indicator can be one of the following: a time indicator, a spatial indicator, or a road network indicator. This represents the mapping relationship between the i-th traffic indicator, the j-th traffic indicator, and the traffic state value output by the traffic state assessment model. This represents the mapping relationship between n traffic indicators and the traffic state values ​​output by the traffic state assessment model;

[0062] Variance decomposition of traffic conditions yields the total variance V(C):

[0063]

[0064] Among them, D i (C) represents the biased variance corresponding to the i-th traffic indicator, and D ij (C) represents the partial variance corresponding to the interaction between the i-th traffic indicator and the j-th traffic indicator, and D 12...n (C) represents the biased variance corresponding to the interaction of the first to nth traffic indicators:

[0065] D i (C)=V(E[C|C i ])

[0066] D ij (C)=V(E[C|C i C j ])-V i -V j

[0067] Where E[] represents the expected calculation, V iV represents the total variance corresponding to the i-th traffic indicator. j Let represent the total variance corresponding to the j-th traffic indicator;

[0068] Through the formula:

[0069]

[0070] Calculate the first-order sensitivity S of the i-th traffic indicator. i ;

[0071] Through the formula:

[0072]

[0073] Calculate the sensitivity of the total effect ST of the i-th traffic indicator. i ;

[0074] Among them, C ～i This represents all traffic indicators except for the i-th traffic indicator;

[0075] If the i-th traffic indicator is a time indicator, then the first-order sensitivity of the i-th traffic indicator and the total effect sensitivity of the i-th traffic indicator are both taken as the time sensitivity of the i-th traffic indicator at the current time.

[0076] If the i-th traffic indicator is a spatial indicator, then the first-order sensitivity of the i-th traffic indicator and the total effect sensitivity of the i-th traffic indicator are both taken as the spatial sensitivity of the i-th traffic indicator at the current time.

[0077] If the i-th traffic indicator is a road network indicator, then the first-order sensitivity of the i-th traffic indicator and the total effect sensitivity of the i-th traffic indicator are both taken as the road network sensitivity of the i-th traffic indicator at the current time.

[0078] Secondly, embodiments of this application provide a traffic indicator sensitivity analysis device, comprising:

[0079] The data acquisition module is used to acquire time index data of multiple time indicators, spatial index data of multiple spatial indicators, and road network index data of multiple road network indicators at T times for the target area; the Tth time is the current time.

[0080] The construction module is used to build a traffic condition assessment model. The traffic condition assessment model includes: a time feature extraction module, a spatial feature extraction module, and a spatiotemporal feature fusion module, a road network feature extraction module, and an assessment module connected in sequence.

[0081] The first feature extraction module uses the time feature extraction module to extract features from all time indicator data to obtain time features, uses the spatial feature extraction module to extract features from all spatial indicator data to obtain spatial features, and uses the spatiotemporal feature fusion module to fuse the time features and spatial features to obtain spatiotemporal fused features.

[0082] The second feature extraction module uses the road network feature extraction module to extract features from the spatiotemporal fusion features and all road network indicator data of the target area to obtain the final traffic features of the target area.

[0083] The evaluation module is used to evaluate the final traffic characteristics and obtain the traffic status of the target area at the current moment.

[0084] The calculation module is used to calculate the time sensitivity of each time indicator, the spatial sensitivity of each spatial indicator, and the road network sensitivity of each road network indicator at the current time, based on the traffic status of the target area at the current time.

[0085] Thirdly, embodiments of this application provide a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the aforementioned traffic indicator sensitivity analysis method.

[0086] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned traffic indicator sensitivity analysis method.

[0087] The above-mentioned solution in this application has the following beneficial effects:

[0088] In the embodiments of this application, time index data of multiple time indicators, spatial index data of multiple spatial indicators, and road network index data of multiple road network indicators at T times are obtained for the target area. Then, a traffic state assessment model is constructed. The time feature extraction module is used to extract features from all time index data to obtain time features. The spatial feature extraction module is used to extract features from all spatial index data to obtain spatial features. The time and spatial features are fused using a spatiotemporal feature fusion module to obtain spatiotemporal fusion features. Then, the road network feature extraction module is used to extract features from the spatiotemporal fusion features and all road network index data of the target area to obtain the final traffic features of the target area. The final traffic features are then evaluated using an evaluation module to obtain the traffic state of the target area at the current time. Finally, based on the traffic state of the target area at the current time, the time sensitivity of each time indicator, the spatial sensitivity of each spatial indicator, and the road network sensitivity of each road network indicator at the current time are calculated. This involves acquiring intelligent connected vehicle data with multiple time and spatial indicators, as well as road network data with road network indicators. It takes into account various factors affecting the traffic status of the target area, improving the accuracy and comprehensiveness of the data. By utilizing the traffic status assessment model, it is possible to conduct in-depth analysis of data from multiple indicators and obtain highly accurate traffic status. Using highly accurate traffic status to obtain the sensitivity of multiple traffic indicators can improve the rationality of the sensitivity.

[0089] Other beneficial effects of this application will be described in detail in the following detailed description section. Attached Figure Description

[0090] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0091] Figure 1 A flowchart of a traffic indicator sensitivity analysis method provided in an embodiment of this application;

[0092] Figure 2 A schematic diagram of traffic indicators provided in one embodiment of this application;

[0093] Figure 3 This is a schematic diagram of the structure of a traffic condition assessment model provided in an embodiment of this application;

[0094] Figure 4 This is a schematic diagram of the structure of a spacetime gating recursive unit provided in an embodiment of this application;

[0095] Figure 5 This is a schematic diagram of the structure of a gated residual attention unit provided in an embodiment of this application;

[0096] Figure 6 A detailed flowchart of a traffic index sensitivity analysis method provided in an embodiment of this application;

[0097] Figure 7 A schematic diagram of the structure of a traffic index sensitivity analysis device provided in an embodiment of this application;

[0098] Figure 8 This is a schematic diagram of the structure of a terminal device provided in an embodiment of this application. Detailed Implementation

[0099] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0100] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0101] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0102] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0103] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0104] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0105] To address the problem that existing analyses of traffic indicators are unreasonable, this application provides a method for traffic indicator sensitivity analysis.

[0106] The following is an illustrative example of the traffic indicator sensitivity analysis method provided in this application.

[0107] like Figure 1 As shown, the traffic indicator sensitivity analysis method provided in this application includes the following steps:

[0108] Step 11: Obtain time index data of multiple time indicators, spatial index data of multiple spatial indicators, and road network index data of multiple road network indicators for the target area at T time points.

[0109] The aforementioned time indicators, spatial indicators, and road network indicators are all traffic indicators that affect the traffic status of the target area. The T-th time is the current time.

[0110] Specifically, for traffic indicators that can be obtained directly (such as the number of intersections, the speed limit on roads, etc.), data on the target area at T time points can be obtained using traffic-related systems or equipment. For traffic indicators that cannot be obtained directly (such as road area ratio, road network density in the area, etc.), existing data can be used for calculation.

[0111] It should be noted that the above time indicators are time-related traffic indicators, while the above spatial indicators are time-independent traffic indicators. Both time and spatial indicators belong to intelligent connected traffic indicators, meaning that the data of these traffic indicators is related to intelligent traffic equipment or systems (such as traffic lights, traffic accident statistics systems, road segment management systems, etc.). The above road network indicators are traffic indicators that reflect the actual situation of roads in the target area, and the above target area is the area where traffic indicator sensitivity analysis needs to be performed.

[0112] For example, 12 traffic indicators are obtained for the target area: road network density, road network connectivity index, number of intersections, road area ratio, road network gradation level, speed limit, traffic light cycle, vehicle travel time, traffic environment index, vehicle operation index, traffic incident index, and travel period. Among these, road network density, road network connectivity index, number of intersections, road area ratio, road network gradation level, and speed limit are all road network indicators; traffic light cycle, vehicle travel time, and travel period are all time indicators; and traffic environment index, vehicle operation index, and traffic incident index are all spatial indicators. For any traffic indicator, its data at T time points {x1, x2, ..., x...} are obtained. T To prevent excessive data from becoming too long when inputting into the traffic state assessment model, the T time points can be divided into multiple sequences. For example, time points 1 to 10 are the first sequence, time points 11 to 20 are the second sequence, and time points 21 to T are the third sequence. In other words, a long data point is divided into three sequences according to the time order.

[0113] The road network density data for the aforementioned area is calculated as the ratio of the sum of all road lengths within the target area to the area of ​​the target area. The road network connectivity index data is calculated as the ratio of twice the sum of the total number of edges in the road network of the target area to the sum of the total number of points in the road network; a higher value indicates stronger connectivity of the road network in that area. The road area ratio is calculated as the ratio of the total area of ​​the regional road network to the area of ​​that region. The road network gradation level is calculated as the sum of the weighted average of the road grade index and length for each road grade. The road grade indices within the area are: expressway = 1, arterial road = 2, secondary arterial road = 3, and local road = 4, reflecting the overall grade level of the regional road network. The intersection number data is obtained by consulting traffic data for the target area.

[0114] The aforementioned traffic light cycle, vehicle running time, and running period are all time indicators. The data for traffic environment index, vehicle operation index, and traffic incident index are obtained by accessing the corresponding intelligent traffic system or equipment. For example, the data for the traffic light cycle is obtained by accessing the traffic light back-end management system.

[0115] It is worth mentioning that by acquiring intelligent connected vehicle data with multiple time and spatial indicators, as well as road network data with road network indicators, the intelligent connected vehicle situation and actual conditions affecting the traffic status of the target area are taken into account, thereby improving the accuracy and comprehensiveness of the data.

[0116] The above traffic indicators will be illustrated with a specific example below.

[0117] like Figure 2As shown, the indicators affecting the regional traffic status of the target area are divided into two categories: road network conditions (i.e., the road network indicators mentioned above) and intelligent connectivity conditions (i.e., the traffic indicators belonging to intelligent connectivity mentioned above). The road network conditions are specifically divided into six traffic indicators: regional road network density, road network connectivity index, number of intersections, road area ratio, road network gradation level, and road speed limit. The intelligent connectivity conditions are specifically divided into six traffic indicators: traffic light cycle, vehicle running time, traffic environment index, vehicle operation index, traffic incident index, and running time segment.

[0118] Step 12: Construct a traffic condition assessment model.

[0119] The traffic condition assessment model mentioned above includes: a time feature extraction module, a spatial feature extraction module, and a spatiotemporal feature fusion module, a road network feature extraction module, and an assessment module connected in sequence.

[0120] Specifically, the temporal feature extraction module includes: a fully convolutional neural network unit and a temporal attention unit connected in sequence;

[0121] The input of the fully convolutional neural network unit is the input of the temporal feature extraction module, and the output of the temporal attention unit is the output of the temporal feature extraction module.

[0122] The spatial feature extraction module includes: a spatial embedding unit and a spatial attention unit connected in sequence;

[0123] The input of the spatial embedding unit is the input of the spatial feature extraction module, and the output of the spatial attention unit is the output of the spatial feature extraction module.

[0124] The aforementioned spatiotemporal feature fusion module includes: a global attention unit and multiple sequentially connected spatiotemporal gating recursive units;

[0125] The input of each spatiotemporal gated recursive unit is connected to the output of the temporal feature extraction module and the output of the spatial feature extraction module. The output of each spatiotemporal gated recursive unit is connected to the input of the global attention unit. The output of the global attention unit is the output of the spatiotemporal feature fusion module.

[0126] The road network feature extraction module includes: a road network embedding unit and a gated residual attention unit;

[0127] The input of the road network embedding unit is connected to all road network index data. The input of the gated residual attention unit is connected to the output of the spatiotemporal feature fusion module and the output of the road network embedding unit. The output of the gated residual attention unit is the output of the road network feature extraction module.

[0128] The assessment module includes: a traffic condition assessment unit;

[0129] The input terminal of the traffic condition assessment unit is the input terminal of the assessment module, and the output terminal of the traffic condition assessment unit is the output terminal of the assessment module.

[0130] It should be noted that the above-mentioned fully convolutional neural network unit is a fully convolutional network for semantic segmentation, the temporal attention unit is a computational unit based on the self-attention mechanism, the above-mentioned spatial embedding unit is an embedding operation, the above-mentioned spatial attention unit is a computational unit based on the multi-head attention mechanism, the above-mentioned global attention unit is a computational unit based on the channel attention mechanism, the above-mentioned spatiotemporal gating recursive unit includes a first gating mechanism layer and a second gating mechanism layer, the above-mentioned road network embedding unit is an embedding operation, and the above-mentioned traffic state assessment unit is a temporal point process (TPP) model without intensity learning strategy.

[0131] The constructed traffic state assessment model needs to be trained. This can be done by using traffic index data and corresponding traffic states from other regions at multiple historical times as a sample set to train the model's parameters. The parameters are continuously changed until the difference between the traffic state output by the model and the traffic state of other regions reaches the expected level. At this point, the traffic state assessment model is considered well-trained. This trained model is then used to assess the traffic state of the target area.

[0132] It is worth mentioning that by utilizing traffic condition assessment models, it is possible to conduct in-depth analysis of data from various indicators and obtain highly accurate traffic conditions.

[0133] The above traffic condition assessment model will be illustrated below with a specific example.

[0134] like Figure 3As shown, the time-series data (i.e., the time index data mentioned above) in the intelligent connected traffic data enters the fully convolutional neural network unit in the time feature extraction module. The output of the fully convolutional neural network unit is connected to the input of the time attention unit. Simultaneously, the spatial sequence data (i.e., the spatial index data mentioned above) in the intelligent connected traffic data enters the spatial embedding unit in the spatial feature extraction module. The output of the spatial embedding unit is connected to the input of the spatial attention unit. The outputs of both the time attention unit and the spatial attention unit are connected to the input of each spatiotemporal gated recurrent unit (STGRU) in the spatiotemporal feature fusion module. The output of each spatiotemporal recurrent unit is connected to the input of the global attention unit. The output of the global attention unit is connected to the gated residual attention network (GRAN) in the road network feature extraction module. The input of the Network (i.e., the gated residual attention unit mentioned above) is connected to the input of the network embedding unit in the network feature extraction module. At the same time, all network index data are input into the network embedding unit in the network feature extraction module. The output of the gated residual attention unit is connected to the input of the traffic state assessment unit in the assessment module. The output of the traffic state assessment unit outputs the traffic state of the target area.

[0135] Step 13: Use the time feature extraction module to extract features from all time indicator data to obtain time features, use the spatial feature extraction module to extract features from all spatial indicator data to obtain spatial features, and use the spatiotemporal feature fusion module to fuse the time features and spatial features to obtain spatiotemporal fused features.

[0136] Specifically, all time indicator data are input into the fully convolutional neural network unit of the time feature extraction module, and then processed using the formula: h t =σ(s t W+b) Calculate shallow features of time h t , where s t Let σ() represent the activation function of the fully convolutional operation, W represent the weight matrix, and b represent the bias. The shallow temporal features are then input into the temporal attention unit, where they undergo a self-attention mechanism to obtain the temporal features. Simultaneously, all spatial indicator data are input into the spatial embedding unit of the spatial feature extraction module, and processed using the formula: h l =Embedding(s l ) Calculate the shallow spatial features h l Where Embedding() represents the embedding strategy, s lAll spatial index data are represented, and then the shallow spatial features are input into the spatial attention unit. After processing by a multi-head attention mechanism, the spatial features are obtained. Then, the aforementioned temporal and spatial features are input into each spatiotemporal gating recursive unit in the spatiotemporal feature fusion module. For each spatiotemporal gating recursive unit, the temporal and spatial features are processed to obtain hidden fused features. All hidden fused features are input into the global attention unit, and after processing by a channel attention mechanism, the stability of the hidden fused features is enhanced, resulting in spatiotemporal fused features.

[0137] It should be noted that the processing steps in the aforementioned spatiotemporal gating recursive unit specifically include:

[0138] The first step is to use the hidden fusion feature H corresponding to the (f-1)th spatiotemporal gated recursive unit. f-1 Two different updates and resets are performed on the time features to obtain the first and second time hidden features corresponding to the f-th spatiotemporal gated recursive unit, respectively.

[0139] Based on the hidden fusion feature H corresponding to the (f-1)th spatiotemporal gated recursive unit f-1 Two different updates and resets are performed on the spatial features to obtain the first spatial hidden feature and the second spatial hidden feature corresponding to the f-th spatiotemporal gated recursive unit, respectively.

[0140] Initial fusion of temporal and spatial features yields initial fused features.

[0141] The second step is to use the formula:

[0142]

[0143]

[0144] Calculate the integrated time features corresponding to the f-th spatiotemporally gated recursive unit.

[0145] Among them, w z The time feature weights are represented by sigmoid(), which represents the activation function. This represents the first-time hidden feature corresponding to the f-th spatiotemporally gated recursive unit. Let f represent the second temporal hidden feature corresponding to the f-th spatiotemporally gated recursive unit, where f = 1, 2, ..., F, and F represents the total number of spatiotemporally gated recursive units. When f = 1, the hidden fusion feature H corresponding to the (f-1)-th spatiotemporally gated recursive unit is... f-1 The value is 0. This represents the weight matrix corresponding to the hidden features in the first space. This represents the weight matrix corresponding to the hidden features in the second space. The symbol ⊙ indicates the bias, and ⊙ indicates the Hadamah product operation.

[0146] Through the formula:

[0147]

[0148]

[0149] Calculate the integrated spatial features corresponding to the f-th spatiotemporally gated recursive unit.

[0150] Among them, w r Represents spatial feature weights, This represents the first spatial hidden feature corresponding to the f-th spatiotemporally gated recursive unit. This represents the second spatial hidden feature corresponding to the f-th spatiotemporally gated recursive unit. This represents the weight matrix corresponding to the hidden features in the first space. This represents the weight matrix corresponding to the hidden features in the second space. Indicates bias;

[0151] The third step is to use the formula:

[0152]

[0153]

[0154] Calculate the hidden fusion feature H corresponding to the f-th spatiotemporally gated recursive unit. f ;

[0155] in, This represents the initial hidden fusion feature corresponding to the Fth spatiotemporal gated recursive unit. Indicates the initial fusion features, This represents the weight matrix corresponding to the initial fused features. This represents the weight matrix of the hidden fusion features corresponding to the (f-1)th spatiotemporally gated recursive unit. Indicates bias;

[0156] Obtain the hidden fusion features H = {H1, H2, ..., H} corresponding to all spatiotemporal gated recursive units. F};

[0157] Where H1 represents the hidden fusion feature corresponding to the first spatiotemporally gated recursive unit, H2 represents the hidden fusion feature corresponding to the second spatiotemporally gated recursive unit, and H... F This represents the hidden fusion feature corresponding to the Fth spatiotemporal gated recursive unit.

[0158] Finally, based on the hidden fusion features H corresponding to all spatiotemporal gated recursive units, the spatiotemporal fusion features are obtained using the global attention unit.

[0159] It should be noted that the aforementioned global attention unit performs feature fusion on the hidden fusion features H corresponding to all spatiotemporal gated recursive units based on the global attention mechanism to obtain spatiotemporal fusion features.

[0160] For example, the above-mentioned hidden fusion feature H based on the (f-1)th spatiotemporal gated recursive unit f-1 Two different updates and resets are performed on the temporal features to obtain the first and second temporal hidden features corresponding to the f-th spatiotemporal gated recursive unit, respectively:

[0161] Through the formula:

[0162]

[0163]

[0164] Obtain the first-time hidden feature corresponding to the f-th spatiotemporal gated recursive unit. Second time hidden features

[0165] Among them, W xr W hr W xz W hz For different weight matrices, b r b z For different biases, x Z H represents the time characteristic. f-1 The hidden fusion feature is the one corresponding to the (f-1)th spatiotemporal gated recursive unit.

[0166] The above-mentioned hidden fusion feature H is based on the (f-1)th spatiotemporal gated recursive unit. f-1 Two different updates and resets are performed on the spatial features to obtain the first and second spatial hidden features corresponding to the f-th spatiotemporal gated recursive unit, respectively:

[0167] Through the formula:

[0168]

[0169]

[0170] Obtain the first spatial hidden feature corresponding to the f-th spatiotemporal gated recursive unit. Second space hidden features

[0171] Where, xR Represents spatial characteristics.

[0172] The above-mentioned spatiotemporal gating recursive unit will be illustrated below with a specific example.

[0173] like Figure 4 As shown, for the f-th spatiotemporally gated recursive unit, the spatial features, temporal features, and the hidden fusion features output by the (f-1)-th spatiotemporally gated recursive unit are input into the f-th spatiotemporally gated recursive unit. Based on the hidden fusion features output by the (f-1)-th spatiotemporally gated recursive unit, the spatial features are updated and reset in two ways to obtain the first spatial hidden feature. Second space hidden features Simultaneously, based on the hidden fusion feature output by the (f-1)th spatiotemporal gated recursive unit, the temporal features are updated and reset in two ways to obtain the first temporal hidden feature. Second time hidden features First Space Hidden Features Second space hidden features Integrated spatial features are obtained after the first gating mechanism layer. Conceal features immediately Second time hidden features The integrated time characteristics are obtained after the second gating mechanism layer. Then based on the integration time characteristics The initial hidden fusion features are calculated by combining the initial fusion features obtained from the fusion of temporal and spatial features. Finally, based on the initial hidden fusion features The hidden fusion feature H corresponding to the f-th spatiotemporally gated recursive unit is calculated. f .

[0174] It is worth mentioning that by extracting features from all time index data and all spatial index data respectively, we can focus on the traffic characteristics of the target area in both time and space. By using the spatiotemporal feature fusion module to interact and fuse the time and spatial features, the resulting spatiotemporal fusion features contain information on both time and space features, which can accurately reflect the traffic status of the target area in time and space.

[0175] Step 14: Use the road network feature extraction module to extract features from the spatiotemporal fusion features and all road network indicator data of the target area to obtain the final traffic features of the target area.

[0176] In some embodiments of this application, the steps for obtaining the final traffic characteristics of the target area include:

[0177] The first step is to extract features from all road network indicator data to obtain road network features.

[0178] Input all road network indicator data into the road network embedding unit using the formula: C' = Embedding(c l The road network features C' are calculated, where Embedding() represents the embedding strategy, and c l This represents all road network indicator data.

[0179] The second step is to use the formula:

[0180] η1=ELU(H * W 1,1 +C'W 1,2 +b1)

[0181] η2=η1W2+b2

[0182]

[0183] Calculate the final traffic characteristics

[0184] Where η1 and η2 both represent intermediate quantities, H * W represents the spatiotemporal fusion characteristics. 1,1 and W 1,2 All represent weight matrices, which can be learned through backpropagation. C' represents road network features. b1 and b2 both represent biases. LN() represents normalization operation. GAU() represents gated layer operation. ELU() represents exponential linear unit activation function. GRAN() represents gated residual attention unit operation.

[0185] It should be noted that the gated residual attention unit in the aforementioned road network feature extraction module includes a first fully convolutional layer, a nonlinear layer, a second fully convolutional layer, an abandonment layer, a gated attention layer, and a norm layer. The input of the first fully convolutional layer is the spatiotemporal fusion feature and the road network feature. The output of the first fully convolutional layer is connected to the input of the second fully convolutional layer, the output of the second fully convolutional layer is connected to the input of the abandonment layer, the output of the abandonment layer is connected to the input of the gated attention layer, and the output of the gated attention layer is connected to the input of the norm layer. Simultaneously, the input of the norm layer is the spatiotemporal fusion feature, and the output of the norm layer is the output of the gated residual attention unit, which outputs the final traffic features. The nonlinear layer uses an exponential linear unit activation function. Both the first and second fully convolutional layers are fully convolutional neural networks. The first convolutional layer is used to further extract features from the spatiotemporal fusion features and road network features. The nonlinear layer is used to nonlinearly transform the processing results of the first convolutional layer to obtain complex features. The second fully convolutional layer is used to further extract features from the complex features. The dropout layer uses the Dropout algorithm to prevent overfitting by discarding some neurons in the second fully convolutional layer, thereby improving the generalization of the road network feature extraction module. The gated attention layer is used to integrate the hidden relationships between spatiotemporal features and further fuse spatiotemporal features. The norm layer is used to prevent overfitting.

[0186] The following example illustrates the gated residual attention unit in the road network feature extraction module.

[0187] like Figure 5 As shown, the spatiotemporal fusion features and road network features are input into the first fully convolutional layer in the gated residual attention unit. After full convolution and exponential linear unit activation (ELU) operations, an intermediate value η1 is obtained. The intermediate value η1 is input into the second fully convolutional layer. After full convolution and a discard layer, an intermediate value η2 is obtained. The intermediate value η2 is input into the gated attention layer. After calculation by the gated attention unit, it is input into the norm layer. It is added to the spatiotemporal fusion features input into the norm layer through residual links. Finally, the traffic features are obtained by norm function Norm operation.

[0188] It is worth mentioning that by using the road network feature extraction module to process the spatiotemporal features and road network features, the final traffic features obtained include the characteristics of the target area in terms of time, space and road network, which can accurately describe the traffic status of the target area over a period of time.

[0189] Step 15: Use the evaluation module to evaluate the final traffic characteristics and obtain the traffic status of the target area at the current moment.

[0190] The evaluation process in the above evaluation module is as follows: the evaluation module is used to calculate the final traffic characteristics to obtain multiple candidate traffic states of the target area and the probability value corresponding to each candidate traffic state. Then, the probability value with the largest value among all probability values ​​is selected as the target probability value, and the candidate traffic state corresponding to the target probability value is taken as the traffic state of the target area.

[0191] It should be noted that the above evaluation module includes a traffic state evaluation unit, which is a Temporal Point Process (TPP) model with a neutral intensity learning strategy. The TPP calculates multiple candidate traffic states for the target area and the probability value corresponding to each candidate traffic state using a log-norm mixed distribution formula.

[0192] For example, the traffic state value ranges from 0 to 20. The traffic state assessment unit calculates the final traffic characteristics of the target area to obtain multiple values ​​of the traffic state and the probability of each value. For example, the probability of the traffic state value being 3 is 20%, the probability of the traffic state value being 6 is 10%, etc. Among them, the probability of the value being 7 is the highest, at 89%. Therefore, the output value of the traffic state of the target area at the current time is 7.

[0193] It is worth mentioning that, based on the data of various traffic indicators of the target area at T times, the traffic state assessment model can comprehensively analyze the traffic state of the target area at the current time, thereby improving the accuracy of traffic state.

[0194] Step 16: Based on the traffic status of the target area at the current time, calculate the time sensitivity of each time indicator, the spatial sensitivity of each spatial indicator, and the road network sensitivity of each road network indicator at the current time.

[0195] Specifically, the traffic conditions described above are as follows:

[0196]

[0197] in, C represents the traffic state value, and C represents the data of all traffic indicators in the target area at time T. Represents a constant term. This represents the mapping relationship between the i-th traffic indicator and the traffic state value output by the traffic state assessment model, where i,j∈{1,2,...,n}, and n represents the total number of traffic indicators. A traffic indicator can be one of the following: a time indicator, a spatial indicator, or a road network indicator. This represents the mapping relationship between the i-th traffic indicator, the j-th traffic indicator, and the traffic state value output by the traffic state assessment model. This represents the mapping relationship between n traffic indicators and the traffic state values ​​output by the traffic state assessment model;

[0198] Variance decomposition of traffic conditions yields the total variance V(C):

[0199]

[0200] Among them, D i (C) represents the biased variance corresponding to the i-th traffic indicator, and D ij (C) represents the partial variance corresponding to the interaction between the i-th traffic indicator and the j-th traffic indicator, and D 12...n (C) represents the biased variance corresponding to the interaction of the first to nth traffic indicators:

[0201] D i (C)=V(E[C|C i ])

[0202] D ij (C)=V(E[C|C i C j ])-V i -V j

[0203] Where E[] represents the expected calculation, V i V represents the total variance corresponding to the i-th traffic indicator. j Let represent the total variance corresponding to the j-th traffic indicator;

[0204] Through the formula:

[0205]

[0206] Calculate the first-order sensitivity S of the i-th traffic indicator. i ;

[0207] Through the formula:

[0208]

[0209] Calculate the sensitivity of the total effect ST of the i-th traffic indicator. i ;

[0210] Among them, C ～i This represents all traffic indicators except for the i-th traffic indicator;

[0211] If the i-th traffic indicator is a time indicator, then the first-order sensitivity of the i-th traffic indicator and the total effect sensitivity of the i-th traffic indicator are both taken as the time sensitivity of the i-th traffic indicator at the current time.

[0212] If the i-th traffic indicator is a spatial indicator, then the first-order sensitivity of the i-th traffic indicator and the total effect sensitivity of the i-th traffic indicator are both taken as the spatial sensitivity of the i-th traffic indicator at the current time.

[0213] If the i-th traffic indicator is a road network indicator, then the first-order sensitivity of the i-th traffic indicator and the total effect sensitivity of the i-th traffic indicator are both taken as the road network sensitivity of the i-th traffic indicator at the current time.

[0214] It should be noted that traffic indicators are classified into levels or categories based on their first-order sensitivity values. The total effect sensitivity reflects the sum of the influence of a system parameter and its interaction with other parameters on the system's output. The calculations described above can be performed using computer software such as Matlab and Mathematica.

[0215] For example, there are 12 traffic indicators. The first-order sensitivity values ​​of these 12 indicators are sorted from largest to smallest. The top three indicators with the highest first-order sensitivity are considered sensitive indicators, the bottom three are considered feasible indicators, and the remaining indicators are considered saturation indicators. Traffic management measures in the target area can be optimized based on the first-order sensitivity, overall effect sensitivity, and the classification of traffic indicators. For example, for sensitive indicators within the area, control or expansionary management methods can be implemented based on their data characteristics to adjust the indicator values, thereby reducing congestion and improving regional traffic capacity. For feasible indicators, related investments can be reduced to decrease government expenditure.

[0216] It is worth mentioning that by acquiring intelligent connected vehicle data with multiple time and spatial indicators, as well as road network data with road network indicators, various factors affecting the traffic status of the target area are taken into account, thereby improving the accuracy and comprehensiveness of the data. By utilizing the traffic status assessment model, in-depth analysis of data from multiple indicators can be performed to obtain highly accurate traffic status data. Using highly accurate traffic status data to obtain the sensitivity of multiple traffic indicators can improve the rationality of the sensitivity.

[0217] The following example illustrates the traffic indicator sensitivity analysis method of this application.

[0218] like Figure 6As shown, the evaluation index dataset consists of two parts. The first part contains data on various traffic indicators from other areas as samples, as well as traffic conditions in other areas. The second part contains data on various traffic indicators from the target area. An evaluation model (i.e., the traffic condition evaluation model mentioned above) is constructed. The model is trained based on the first part of the evaluation index dataset to obtain the optimal model parameters (i.e., the trained traffic condition evaluation model). The parameter value range of the data in the second part is confirmed (i.e., data preprocessing is performed, such as removing data that exceeds the value range and dividing redundant data into multiple sequences). Then, random sampling is performed to generate the target dataset (i.e., the various traffic indicator data of the target area mentioned above). The target dataset is input into the trained traffic condition evaluation model, and the response is output. The variance of the output result is decomposed to obtain the sensitivity.

[0219] The following is an exemplary description of the traffic indicator sensitivity analysis device provided in this application.

[0220] like Figure 7 As shown in the figure, this application provides a traffic indicator sensitivity analysis device 700, which includes:

[0221] The data acquisition module 701 is used to acquire time index data of multiple time indicators, spatial index data of multiple spatial indicators, and road network index data of multiple road network indicators at T times for the target area; the Tth time is the current time.

[0222] Module 702 is used to construct a traffic state assessment model. The traffic state assessment model includes: a time feature extraction module, a spatial feature extraction module, and a spatiotemporal feature fusion module, a road network feature extraction module, and an assessment module connected in sequence.

[0223] The first feature extraction module 703 uses the time feature extraction module to extract features from all time indicator data to obtain time features, uses the spatial feature extraction module to extract features from all spatial indicator data to obtain spatial features, and uses the spatiotemporal feature fusion module to fuse the time features and spatial features to obtain spatiotemporal fused features.

[0224] The second feature extraction module 704 uses the road network feature extraction module to extract features from the spatiotemporal fusion features and all road network indicator data of the target area to obtain the final traffic features of the target area.

[0225] Evaluation module 705 is used to evaluate the final traffic characteristics and obtain the traffic status of the target area at the current time.

[0226] The calculation module 706 is used to calculate the time sensitivity of each time indicator, the spatial sensitivity of each spatial indicator, and the road network sensitivity of each road network indicator at the current time, based on the traffic status of the target area at the current time.

[0227] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0228] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0229] like Figure 8 As shown, an embodiment of this application provides a terminal device, wherein the terminal device D10 of this embodiment includes: at least one processor D100 ( Figure 8 The diagram shows only one processor, a memory D101, and a computer program D102 stored in the memory D101 and executable on the at least one processor D100, wherein the processor D100 executes the computer program D102 to implement the steps in any of the above method embodiments.

[0230] Specifically, when the processor D100 executes the computer program D102, it acquires time index data of multiple time indicators, spatial index data of multiple spatial indicators, and road network index data of multiple road network indicators at T times for the target area. Then, it constructs a traffic state assessment model. Next, it uses a time feature extraction module to extract features from all time index data to obtain time features, and a spatial feature extraction module to extract features from all spatial index data to obtain spatial features. Then, it uses a spatiotemporal feature fusion module to fuse the time features and spatial features to obtain spatiotemporal fusion features. Then, it uses a road network feature extraction module to extract features from the spatiotemporal fusion features and all road network index data of the target area to obtain the final traffic features of the target area. Then, it uses an evaluation module to evaluate the final traffic features to obtain the traffic state of the target area at the current time. Finally, based on the traffic state of the target area at the current time, it calculates the time sensitivity of each time indicator, the spatial sensitivity of each spatial indicator, and the road network sensitivity of each road network indicator at the current time. This involves acquiring intelligent connected vehicle data with multiple time and spatial indicators, as well as road network data with road network indicators. It takes into account various factors affecting the traffic status of the target area, improving the accuracy and comprehensiveness of the data. By utilizing the traffic status assessment model, it is possible to conduct in-depth analysis of data from multiple indicators and obtain highly accurate traffic status. Using highly accurate traffic status to obtain the sensitivity of multiple traffic indicators can improve the rationality of the sensitivity.

[0231] The processor D100 can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0232] In some embodiments, the memory D101 may be an internal storage unit of the terminal device D10, such as a hard disk or memory of the terminal device D10. In other embodiments, the memory D101 may be an external storage device of the terminal device D10, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the terminal device D10. Furthermore, the memory D101 may include both internal and external storage units of the terminal device D10. The memory D101 is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory D101 can also be used to temporarily store data that has been output or will be output.

[0233] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above.

[0234] This application provides a computer program product that, when run on a terminal device, enables the terminal device to implement the steps described in the various method embodiments above.

[0235] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying the computer program code to the traffic indicator sensitivity analysis method device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0236] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0237] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0238] The above description is the preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principles described in this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for sensitivity analysis of traffic indicators, characterized in that, include: Obtain multiple time indicators for the target region T Time-based data and multiple spatial indicators at a given moment T Spatial indicator data and multiple road network indicators at a given time point T Road network index data at time 1; T The current moment is represented by the current moment. A traffic condition assessment model is constructed; the traffic condition assessment model includes: a time feature extraction module, a spatial feature extraction module, and a spatiotemporal feature fusion module, a road network feature extraction module, and an assessment module connected in sequence. The time feature extraction module is used to extract features from all time indicator data to obtain time features. The spatial feature extraction module is used to extract features from all spatial indicator data to obtain spatial features. The spatiotemporal feature fusion module is used to fuse the time features and spatial features to obtain spatiotemporal fused features. The road network feature extraction module is used to extract features from the spatiotemporal fusion features and all road network indicator data of the target area to obtain the final traffic features of the target area. The final traffic characteristics are evaluated using the evaluation module to obtain the traffic status of the target area at the current moment; Based on the traffic status of the target area at the current time, calculate the time sensitivity of each time indicator, the spatial sensitivity of each spatial indicator, and the road network sensitivity of each road network indicator at the current time. The step of calculating the time sensitivity of each time indicator, the spatial sensitivity of each spatial indicator, and the road network sensitivity of each road network indicator at the current time, based on the traffic status of the target area at the current time, includes: The traffic situation is as follows: in, This represents the traffic state value. This indicates that all traffic indicators in the target area are... T Data at any given moment Represents a constant term. Indicates the first The mapping relationship between each traffic indicator and the traffic state value output by the traffic state assessment model. , This represents the total number of the traffic indicators, which are one of the time indicators, spatial indicators, and road network indicators. Indicates the first The traffic indicator and the first The mapping relationship between each traffic indicator and the traffic state value output by the traffic state assessment model. express The mapping relationship between each traffic indicator and the traffic state value output by the traffic state assessment model; The traffic conditions are decomposed using variance analysis to obtain the total variance of the traffic conditions. : in, Indicates the first The skewness of each traffic indicator Indicates the first The traffic indicator and the first The partial variance corresponding to the interaction between traffic indicators express The skewness of the interaction between the traffic indicators: in, This indicates the expected calculation. Indicates the first The total variance corresponding to each traffic indicator Indicates the first The total variance corresponding to each traffic indicator; Through the formula: Calculate the first First-order sensitivity of traffic indicators ; Through the formula: Calculate the first Sensitivity of the total effect of individual traffic indicators ; in, Indicates except the first All other traffic indicators besides the traffic indicator; If the first If the traffic indicator is a time indicator, then the first one will be... The first-order sensitivity of the traffic indicator and the aforementioned... The overall effect sensitivity of each traffic indicator is used as the first... The time sensitivity of a traffic indicator at the current moment; If the first If the first traffic indicator is a spatial indicator, then the first... The first-order sensitivity of the traffic indicator and the aforementioned... The overall effect sensitivity of each traffic indicator is used as the first... The spatial sensitivity of a traffic indicator at the current moment; If the first If the first traffic indicator is a road network indicator, then the first... The first-order sensitivity of the traffic indicator and the aforementioned... The overall effect sensitivity of each traffic indicator is used as the first... The sensitivity of a traffic indicator to the road network at the current moment.

2. The traffic indicator sensitivity analysis method according to claim 1, characterized in that, The time feature extraction module includes: A fully convolutional neural network unit and a temporal attention unit connected in sequence; The input terminal of the fully convolutional neural network unit is the input terminal of the temporal feature extraction module, and the output terminal of the temporal attention unit is the output terminal of the temporal feature extraction module; The spatial feature extraction module includes: a spatial embedding unit and a spatial attention unit connected in sequence; The input terminal of the spatial embedding unit is the input terminal of the spatial feature extraction module, and the output terminal of the spatial attention unit is the output terminal of the spatial feature extraction module.

3. The traffic indicator sensitivity analysis method according to claim 1, characterized in that, The spatiotemporal feature fusion module includes: A global attention unit and multiple sequentially connected spatiotemporally gated recursive units; The input of each spatiotemporal gated recursive unit is connected to the output of the temporal feature extraction module and the output of the spatial feature extraction module. The output of each spatiotemporal gated recursive unit is connected to the input of the global attention unit. The output of the global attention unit is the output of the spatiotemporal feature fusion module. The road network feature extraction module includes: a road network embedding unit and a gated residual attention unit; The input of the road network embedding unit is connected to all road network index data. The input of the gated residual attention unit is connected to the output of the spatiotemporal feature fusion module and the output of the road network embedding unit. The output of the gated residual attention unit is the output of the road network feature extraction module. The assessment module includes: a traffic condition assessment unit; The input terminal of the traffic condition assessment unit is the input terminal of the assessment module, and the output terminal of the traffic condition assessment unit is the output terminal of the assessment module.

4. The traffic indicator sensitivity analysis method according to claim 3, characterized in that, The process of fusing the temporal and spatial features using the spatiotemporal feature fusion module to obtain spatiotemporal fused features includes: Based on the Hidden fusion features corresponding to each spatiotemporal gated recursive unit The time feature is updated and reset in two different ways to obtain the first time. The first and second temporal hidden features corresponding to each spatiotemporal gated recursive unit; Based on the first Hidden fusion features corresponding to each spatiotemporal gated recursive unit The spatial features are updated and reset in two different ways to obtain the first... The first and second spatial hidden features corresponding to each spatiotemporal gated recursive unit; The temporal and spatial features are initially fused to obtain initial fused features; Through the formula: Calculate the first The integrated time characteristics corresponding to each spatiotemporal gated recursive unit ; in, Represents the weight of time features. This represents the activation function. Indicates the first The first-time hidden features corresponding to each spatiotemporal gated recursive unit Indicates the first The second temporal hidden feature corresponding to each spatiotemporal gated recursive unit , This represents the total number of the spatiotemporally gated recursive units, when At that time, the first Hidden fusion features corresponding to each spatiotemporal gated recursive unit The value is 0. This represents the weight matrix corresponding to the first spatial hidden features. This represents the weight matrix corresponding to the hidden features in the second space. Indicates bias. This represents the Hadama product operation; Through the formula: Calculate the first Integrated spatial characteristics corresponding to each spatiotemporal gated recursive unit ; in, Represents spatial feature weights, Indicates the first The first spatial hidden feature corresponding to each spatiotemporal gated recursive unit Indicates the first The second spatial hidden features corresponding to each spatiotemporal gated recursive unit This represents the weight matrix corresponding to the first spatial hidden features. This represents the weight matrix corresponding to the hidden features in the second space. Indicates bias; Through the formula: Calculate the first Hidden fusion features corresponding to each spatiotemporal gated recursive unit ; in, Indicates the first The initial hidden fusion features corresponding to each spatiotemporal gated recursive unit This represents the initial fusion feature. This represents the weight matrix corresponding to the initial fused features. Indicates the first The weight matrix of the hidden fusion features corresponding to each spatiotemporally gated recursive unit. Indicates bias; Obtain the hidden fusion features corresponding to all spatiotemporal gated recursive units. ; in, Indicates the first Hidden fusion features corresponding to each spatiotemporal gated recursive unit Indicates the first Hidden fusion features corresponding to each spatiotemporal gated recursive unit Indicates the first Hidden fusion features corresponding to each spatiotemporally gated recursive unit; Based on the hidden fusion features corresponding to all spatiotemporal gated recursive units The spatiotemporal fusion features are obtained using the global attention unit.

5. The traffic indicator sensitivity analysis method according to claim 1, characterized in that, The step of using the road network feature extraction module to extract features from the spatiotemporal fusion features and all road network indicator data of the target area to obtain the final traffic features of the target area includes: Feature extraction is performed on all the road network indicator data to obtain road network features; Through the formula: Calculate the final traffic characteristics ; in, and Both represent intermediate quantities. This indicates the spatiotemporal fusion feature. and Both represent weight matrices. This indicates the characteristics of the road network. and Both represent bias. This indicates a normalization operation. This indicates gating layer operations. This represents the exponential linear unit activation function.

6. The traffic indicator sensitivity analysis method according to claim 1, characterized in that, The step of evaluating the final traffic characteristics using the evaluation module in the traffic state assessment model to obtain the traffic state of the target area at the current time includes: The evaluation module is used to calculate the final traffic characteristics to obtain multiple candidate traffic states for the target area and the probability value corresponding to each candidate traffic state. The probability value with the largest value among all probability values ​​is selected as the target probability value, and the candidate traffic state corresponding to the target probability value is taken as the traffic state of the target area.

7. A traffic indicator sensitivity analysis device, characterized in that, include: The data acquisition module is used to acquire multiple time indicators for the target area. T Time-based data and multiple spatial indicators at a given moment T Spatial indicator data and multiple road network indicators at a given time point T Road network index data at time 1; T The current moment is represented by the current moment. A construction module is used to construct a traffic state assessment model; the traffic state assessment model includes: a time feature extraction module, a spatial feature extraction module, and a spatiotemporal feature fusion module, a road network feature extraction module, and an assessment module connected in sequence. The first feature extraction module uses the time feature extraction module to extract features from all time indicator data to obtain time features, uses the spatial feature extraction module to extract features from all spatial indicator data to obtain spatial features, and uses the spatiotemporal feature fusion module to fuse the time features and spatial features to obtain spatiotemporal fused features. The second feature extraction module uses the road network feature extraction module to extract features from the spatiotemporal fusion features and all road network index data of the target area to obtain the final traffic features of the target area. An evaluation module is used to evaluate the final traffic characteristics to obtain the traffic status of the target area at the current moment. The calculation module is used to calculate the time sensitivity of each time indicator, the spatial sensitivity of each spatial indicator, and the road network sensitivity of each road network indicator at the current time, based on the traffic status of the target area at the current time. Specifically, the calculation module is used to implement: The traffic situation is as follows: in, This represents the traffic state value. This indicates that all traffic indicators in the target area are... T Data at any given moment Represents a constant term. Indicates the first The mapping relationship between each traffic indicator and the traffic state value output by the traffic state assessment model. , This represents the total number of the traffic indicators, which are one of the time indicators, spatial indicators, and road network indicators. Indicates the first The traffic indicator and the first The mapping relationship between each traffic indicator and the traffic state value output by the traffic state assessment model. express The mapping relationship between each traffic indicator and the traffic state value output by the traffic state assessment model; The traffic conditions are decomposed using variance analysis to obtain the total variance of the traffic conditions. : in, Indicates the first The skewness of each traffic indicator Indicates the first The traffic indicator and the first The partial variance corresponding to the interaction between traffic indicators express The skewness of the interaction between the traffic indicators: in, This indicates the expected calculation. Indicates the first The total variance corresponding to each traffic indicator Indicates the first The total variance corresponding to each traffic indicator; Through the formula: Calculate the first First-order sensitivity of traffic indicators ; Through the formula: Calculate the first Sensitivity of the total effect of individual traffic indicators ; in, Indicates except the first All other traffic indicators besides the traffic indicator; If the first If the traffic indicator is a time indicator, then the first one will be... The first-order sensitivity of the traffic indicator and the aforementioned... The overall effect sensitivity of each traffic indicator is used as the first... The time sensitivity of a traffic indicator at the current moment; If the first If the first traffic indicator is a spatial indicator, then the first... The first-order sensitivity of the traffic indicator and the aforementioned... The overall effect sensitivity of each traffic indicator is used as the first... The spatial sensitivity of a traffic indicator at the current moment; If the first If the first traffic indicator is a road network indicator, then the first... The first-order sensitivity of the traffic indicator and the aforementioned... The overall effect sensitivity of each traffic indicator is used as the first... The sensitivity of a traffic indicator to the road network at the current moment.

8. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the traffic indicator sensitivity analysis method as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the traffic indicator sensitivity analysis method as described in any one of claims 1 to 6.

Images