Intelligent traffic prediction control method and system based on dynamic functional cell body coupling

Abstract

The invention belongs to the technical field of intelligent traffic, particularly relates to an intelligent traffic prediction control method and system based on dynamic functional cell body coupling, and aims to solve the problems that an existing traffic planning method is low in prediction accuracy and poor in adaptability in a dynamic network. According to the method, high-order structure features of a traffic network are extracted based on directed motif and a graph neural network (GNN), and urban functional cell bodies are dynamically divided; continuous evolution of the urban functional cell body structure is ensured through a time smoothing mechanism; the mutual feedback intensity between the urban functional cell bodies is predicted by utilizing an urban functional cell body embedding and time sequence model (such as LSTM); and a prediction result is adaptively adjusted in combination with dynamic states (birth, expansion, shrinkage and death) of urban functional cell bodies. On the basis of dynamically adapting to a time-varying traffic environment, regional traffic flow prediction is realized, and when the method is applied to traffic planning, regional flow management, congestion early warning and infrastructure planning can be optimized, and intelligent traffic and green city development are promoted.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent transportation technology and relates to a traffic prediction and control system that integrates the dynamic mutual feedback mechanism of urban spatial organization and a time-series graph neural network. The present invention is particularly suitable for traffic management in large cities with significant human activity rhythmic characteristics. Specifically, it relates to an intelligent traffic prediction and control method and system based on dynamic functional cell coupling. Background Art

[0002] With the continuous improvement of urbanization, the operational efficiency of transportation systems has become one of the most important indicators for measuring a city's comprehensive governance capabilities and the quality of life of residents. Traffic congestion has long plagued urban development, especially in large and medium-sized cities, with typical phenomena such as periodic peak congestion, structural flow imbalances, and abnormal congestion induced by emergencies. Traditional traffic control methods, such as fixed signal timing strategies and induction models based on empirical rules, generally rely on static data or manually set control logic, making it difficult to meet the high time-variability, high complexity, and multi-scale linkage requirements of modern urban transportation networks. Therefore, improving the intelligence, dynamism, and regional adaptability of urban traffic prediction and control systems has become a core goal of intelligent transportation system (ITS) research.

[0003] In recent years, with the rapid development of the Internet of Things, big data, and artificial intelligence technologies, traffic flow prediction methods based on deep learning have made significant progress in accuracy and generalization. In particular, models such as Graph Neural Networks (GNN) and Long Short-Term Memory (LSTM) have been widely used in time series prediction tasks for complex traffic flows due to their excellent spatiotemporal modeling capabilities. However, despite existing technologies, such as the Chinese patent publication number CN118675324A, which discloses a traffic flow prediction system based on deep learning and dynamic network analysis and its application method, a traffic flow prediction system combining Internet of Things technology, graph convolution mechanism, and time series prediction network is proposed. The system perceives multi-source traffic data, constructs a spatiotemporal feature matrix that reflects the evolution law of the traffic network, and uses the graph attention mechanism and time series model for training to achieve the prediction of traffic flow evolution trends. This system integrates multiple algorithmic frameworks, including the Graph Attention Mechanism and Convolutional Neural Networks (CNNs), to improve its ability to depict changing traffic flow patterns. This, to a certain extent, addresses the shortcomings of traditional static models in dealing with dynamic traffic scenarios. However, it still has significant limitations in the following areas, making it difficult to fully adapt to the evolving trends of modern urban transportation systems:

[0004] First: Insufficient consideration of spatial organization:

[0005] Existing deep learning technologies, when used for traffic forecasting and management, rely heavily on surface-level correlations in historical traffic data and direct connections between nodes in a graph structure, neglecting the structural element of urban spatial organization that has a long-term impact on traffic evolution: urban spatial organization. In real cities, the spatial distribution of traffic flow is closely related to the layout of urban functional areas. Different functional areas have stable or cyclical traffic patterns due to differences in population density, land use, and economic activity. As a core planning dimension of intelligent transportation systems, urban spatial organization essentially reflects the spatial configuration and collaborative relationships of urban functional modules. The rigid time-slot division strategy adopted by traditional traffic control systems is difficult to respond to the multi-scale complexity of modern urban transportation networks and the time-varying coupling effects that emerge between functional units. Existing methods fail to incorporate the spatial coupling relationship system of urban functional modules into the modeling system, resulting in a lack of explanatory power of spatial organization in the prediction results at the macro level.

[0006] Second: Lack of dynamic feedback mechanism:

[0007] Existing deep learning technologies, when modeling interactions between traffic nodes, fail to capture the dynamic feedback loops between functional regions. In real urban traffic networks, human activities exhibit significant temporal pulsation, such as commuting tides and leisure travel waves, and traffic interactions between different regions exhibit nonlinear, time-varying coupling characteristics. For example, commuting tidal flows between residential and work areas, and weekend traffic flows between commercial and tourist areas, all exhibit distinct rhythmicity, predictability, and mutual driving. Existing methods fail to identify this macroscopic feedback loop stemming from the laws of human activity. Instead, they simplify interregional traffic relationships into isolated edge weight evolutions, limiting the model's ability to reflect interregional traffic linkages, impacting both sensitivity to abnormal traffic fluctuations and prediction accuracy.

[0008] Furthermore, urban transportation networks often undergo structural changes due to emergencies (such as traffic accidents, extreme weather, and large-scale events) or functional area adjustments (such as new district construction and road reconstruction). These changes can render existing traffic patterns ineffective. Existing traffic forecasting methods using deep learning technology primarily rely on fixed model parameters constructed during training. These methods lack online updating and structural adjustment mechanisms, making it difficult to achieve real-time response to sudden changes in the traffic network and adaptively adjust forecasting strategies, thus compromising the timeliness and effectiveness of forecasts.

[0009] Therefore, how to effectively integrate the temporal characteristics of human activities with traffic dynamics to achieve accurate prediction and adaptive control is a technical challenge that needs to be overcome urgently. Summary of the Invention

[0010] In view of this, the present invention aims to solve the technical problem that traditional traffic control means are difficult to meet the high time-variability, high complexity and multi-scale linkage requirements of modern urban traffic networks, and provides an intelligent traffic prediction and control method and system based on dynamic functional cell coupling. This method implicitly divides the city into several urban functional cells based on vehicle trajectory data. By constructing a dynamic urban network, extracting motif features, dividing urban functional cells, and predicting the changes in the mutual feedback intensity between cells, the mutual feedback intensity between functional cells is calculated to predict traffic changes between functional cells, thereby realizing accurate prediction and adaptive control of urban traffic, and achieving the effect of improving the operating efficiency of the traffic system and reducing congestion.

[0011] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0012] The first object of the present application is to disclose an intelligent traffic prediction control system and method based on dynamic functional cell coupling, comprising the following steps:

[0013] S1: Obtain trajectory data of taxis, shared cars, and other vehicles, and construct a static urban network at the initial moment using the taxi area as the main geographical unit;

[0014] S2: Based on the static network, a dynamic urban network is constructed with a given time interval Δ;

[0015] S3: By extracting motif features and using the GNN model to aggregate the dynamic embedding of geographic locations, dynamic urban cell structures are identified;

[0016] S4: Based on the dynamic urban cell structure, the mutual feedback characteristics between dynamic urban cells are extracted based on the LSTM model to predict the changes in the mutual feedback intensity of urban cells;

[0017] S5: Use prediction results to guide urban transportation planning.

[0018] The S1 is specifically as follows:

[0019] Initial static city network G0 construction: The trajectory data of taxis, shared cars, etc. are collected, and the taxi area is used as the main geographical unit as the node set V0 in the city network. The driving trajectories of all vehicles are used as supplementary directed edges E0 in the city network, and the driving trajectories between regions are constructed to construct the static city network G0 =<V0,E0> .

[0020] The S2 is specifically as follows:

[0021] Constructing a dynamic urban network G: Given a time interval Δ, 24 hours a day can be divided into a set of discrete time intervals {Δ, 2Δ, ..., tΔ}. Based on this set, the dynamic urban network can be represented as a set of city snapshots at multiple time intervals G = {G0, G1, ..., G t The static urban network G at the tth time interval t = <V t ,E t >, where V t , E t They represent the city network node set and edge set at the tth time interval. Each node Represents various geographical locations in the city, each edge Represents the interaction between geographical locations (if the cells are reachable, there is an edge; if they are not reachable, there is no edge). Each edge has a weight represents the node in the city network at the tth time interval and At the same time, G t It is composed of multiple city cells. Because the urban network is changing dynamically, the urban cell also changes dynamically. All include G t A set of nodes and edges in the cell body, and the interaction strength inside the cell body is relatively strong, while the interaction strength outside the cell body (i.e., between the cell bodies) is relatively weak.

[0022] The S3 is specifically as follows:

[0023] The dynamic city cell structure at each moment is realized through the following steps Identification:

[0024] Step 1: Dynamic motif feature extraction.

[0025] (1) For each node Calculate the number of motifs it participates in using the following formula:

[0026]

[0027] Where A is G t The directed weighted adjacency matrix of

[0028] (2) For the static city network G0, obtain the set of all motif instances in the network For each node Get the motif instance M it participates in and calculate Weighted motif strength:

[0029]

[0030] in, is a node The product of the weights of all edges in M, the temporally smoothed motif strength where α is the first smoothing factor,

[0031] (3) Constructing the weighted motif matrix W M , Representation node The weighted connection strength in a directed motif is used to capture The association strength in the directed weighted motif enhances the recognition of the urban cell structure. The formula is as follows:

[0032]

[0033] To avoid weight scale differences, normalize W M , the formula is as follows:

[0034]

[0035] in Normalized W M Facilitates GNN aggregation and balances the contributions of different nodes.

[0036] Time-smoothed motif matrix:

[0037]

[0038] γ is the second smoothing factor.

[0039] In order to reduce the time complexity of directed weighted motif feature extraction, a motif sampling algorithm based on random walk is used to approximate the N m (v) and

[0040] Step 2: GNN feature aggregation:

[0041] Use GNN to aggregate node features, combined with the directed weighted adjacency matrix A and the weighted motif matrix W M Generate node embeddings.

[0042] (1) Initialization features: node features Including out-degree, in-degree, and smooth motif strength

[0043] (2) GNN aggregation:

[0044]

[0045] in I is the identity matrix, which retains the node's own information by adding self-loops. (l) is the node embedding of layer l, are learnable weights and σ is the activation function. yes The out-degree matrix is a combination matrix used to fuse edge weight information and motif information. λ∈[0,1] is a hyperparameter.

[0046] On this basis, we get the embedded smooth Z t , the formula is as follows:

[0047]

[0048] Where β∈[0,1] is the third smoothing factor.

[0049] Step 3: Identification and adjustment of dynamic urban cells.

[0050] Using the geometric distance and node weights in the embedding space, a weighted K-means method is used to identify dynamic urban cells. The centroid of the urban functional cell at time t-1 is used as a reference when initializing the cluster center. The optimization objectives are as follows:

[0051]

[0052] in, is a node The weight of μ, combining motif strength and degree. t,i It is the functional cell of the city The weighted centroid of is as follows:

[0053]

[0054] When t=0, randomly initialize k0 centroids. If t>=1, use The center of mass μ t-1,i As the initial value, add or delete the centroid to match k t .

[0055] By calculating the weighted similarity between dynamic city cells, we can identify and cell body state.

[0056]

[0057] Dynamic city cell birth: if With all of but For the new urban cell;

[0058] Dynamic city cell death: if With all of but die;

[0059] Dynamic city cell continues: if and satisfy and but yes Continuation of

[0060] Dynamic urban cell expansion: If and but merge

[0061] Dynamic city cell shrinks: if and The city cell shrinks.

[0062] The S4 is specifically as follows:

[0063] (1) Cell level feature extraction: After obtaining the node embedding Z t Based on the The embedding formula is as follows:

[0064]

[0065] (2) Constructing cell-body mutual feedback characteristics

[0066] First, for each cell body Constructing feature vectors The formula is as follows:

[0067]

[0068] in, is the sum of the motif intensities of all nodes inside the cell body, are the sum of the out-degree and in-degree of the nodes inside the cell body respectively;

[0069] Then, construct the mutual feedback feature between cell pairs The formula is as follows:

[0070]

[0071] in Indicates that at time t, the cell body The formula for the mutual feedback strength between is as follows:

[0072]

[0073] in, It's the edge The weight at time t.

[0074] (3) Timing Modeling

[0075] Use LSTM to model the time series of the mutual feedback strength of urban cell pairs and input historical mutual feedback features The LSTM model is constructed as follows:

[0076] h t ,c t =LSTM(X i,j ,h t-1 ,c t-1 )

[0077]

[0078] where h t ,c t is the hidden state and cell state of LSTM; W out is the weight of the output layer, b out is the output layer bias.

[0079] The loss function is:

[0080]

[0081] N is the number of city cell pairs.

[0082] By calculating the weighted similarity between dynamic city cells, we can identify and The cell state can be used to dynamically adjust the mutual feedback strength of the urban cell body.

[0083]

[0084] Dynamic city cell birth: if is a newly born cell body, which is initialized based on the average mutual feedback strength of the current time step to predict the mutual feedback strength of the new cell body:

[0085]

[0086] Dynamic city cell death: if Died at time t+h,

[0087] Dynamic city cell expansion and contraction: If and Match, adjust for scale changes:

[0088]

[0089] in is the basic mutual feedback strength of the city functional cell pair at that moment,

[0090] is the predicted value of interaction strength.

[0091] Compared with the existing technology, the intelligent traffic prediction control method and system based on dynamic functional cell coupling described in the present invention has the following advantages:

[0092] (1) The present invention captures the directed weighted high-order structure of the traffic network (such as the cyclic flow pattern reflected by the directed triangle) through the motif-enhanced GNN, extracts regional-level features based on the dynamic urban functional cells, and combines the LSTM time series model to predict the mutual feedback intensity between urban functional cells (such as the vehicle flow between regions). It can more accurately predict the flow between urban functional cells, support more scientific signal light optimization, route planning and congestion warning, reduce traffic delays, and improve urban traffic efficiency.

[0093] (2) This method introduces a temporal smoothing mechanism (including smoothing of motif strength, node embedding, and interaction strength) to ensure the continuous evolution of urban functional cell structure and interaction strength. At the same time, by incremental motif counting and reusing GNN weights, the urban functional cell identification and prediction results are dynamically updated. Dynamic adaptability supports real-time traffic management, such as quickly responding to traffic changes caused by new regional development or emergencies (such as large-scale events) and optimizing traffic resource allocation.

[0094] (3) This method clearly identifies the dynamic state of traffic communities (such as the "birth" of new commercial districts and the "expansion" of main road communities during peak hours) through weighted similarity matching and state determination algorithms, and adaptively adjusts the interaction intensity prediction based on the state. This can support traffic planners to adjust strategies in a timely manner, such as planning new bus routes for new urban functional cells or increasing road capacity for expanding urban functional cells, thereby promoting the sustainable development of urban transportation networks. BRIEF DESCRIPTION OF THE DRAWINGS

[0095] Figure 1 is a flow chart of an intelligent traffic prediction and control method based on dynamic functional cell coupling according to an embodiment of the present invention;

[0096] Figure 2 Schematic diagram of the structure of the urban functional cell mutual feedback time series prediction system in an embodiment of the present invention. DETAILED DESCRIPTION

[0097] The following will be combined with the accompanying drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field are within the scope of protection of this application.

[0098] It should be noted that, in the present application, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the statement "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in the opposite order according to the functions involved. For example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0099] Existing traffic control systems often rely on rigid time-segmentation strategies, making them incapable of addressing the multiscale complexity of modern urban transportation networks and the time-varying coupling effects between functional units. With the development of ubiquitous sensing technology, the acquisition of big data on human activity has provided a new methodological foundation for quantifying and analyzing the dynamic characteristics of urban functional cells, making it possible to quantify and analyze the dynamic characteristics of urban functional cells. However, existing traffic control systems fail to incorporate the nonlinear, time-varying relationships between functional cells into the control decision-making framework, resulting in a lack of dynamic feedback loops. Therefore, intelligent traffic prediction systems based on dynamic functional cells and time-series graph neural networks have emerged, offering new insights and solutions to address urban traffic issues.

[0100] like Figures 1-2 As shown, the present application discloses an intelligent traffic prediction control method based on dynamic functional cell coupling, comprising:

[0101] S1: Obtain urban vehicle trajectory data, use taxi areas as the main geographical units, and construct a static urban network at the initial moment. The nodes of the static urban network represent the geographical locations of the cities, and the edges represent the traffic interaction relationships between nodes, with directions and weights.

[0102] S2: Constructing a static urban network into a dynamic urban network, wherein the dynamic urban network is composed of urban network snapshots at multiple time segments, and is used to characterize the time-varying structure of the urban transportation network;

[0103] S3: In the network snapshot of each time segment, by extracting motif features, constructing a motif feature matrix, and using the GNN model to aggregate the dynamic embedding of geographic locations, the city functional cells are divided, and combined with the cell state at the previous moment, the birth, death, expansion, contraction, or continuous state of the cell body is identified;

[0104] S4: Based on the dynamic urban functional cell structure, the mutual feedback feature vectors between urban functional cells are constructed to form the mutual feedback time series between cell pairs. The long short-term memory network (LSTM) is used to model the mutual feedback time series between cell pairs to predict the changes in the mutual feedback intensity between urban functional cells in subsequent time segments.

[0105] S5: Provide support for urban traffic planning based on the prediction results, including any one or more of signal light timing, route navigation, traffic warning and regional scheduling suggestions.

[0106] The intelligent traffic prediction and control method described in the present application collects driving trajectory data of vehicles such as taxis and shared vehicles, and uses taxi driving areas as the main reference to divide the city into several taxi operating areas to construct an initial static traffic network. In this model, each node represents a specific location in the city, and the edges between nodes represent the traffic flow relationship between these locations. These edges are not only directional but also have weights to represent the size of the traffic flow. The static network is then split into a series of network snapshots that can reflect time-varying characteristics according to time intervals. Each snapshot reflects the urban traffic conditions at a specific time point, thereby capturing the time-varying characteristics of the traffic network. In the network snapshot of each time segment, by analyzing the specific pattern (m otif features), and using graph neural network (GNN) technology, the dynamic characteristics of geographic location are integrated into the model to divide different urban functional cells. These cells represent areas in the city with similar functions or traffic characteristics. At the same time, combined with the cell state at the previous moment, the state of urban functional cells is updated to identify the addition, disappearance and structural changes (expansion, contraction or continuity) of cells. Based on these dynamically changing functional cell structures, the mutual feedback strength between each pair of functional cells is counted according to the division results and a mutual feedback time series is generated. Finally, this sequence is input into the LSTM model for time series prediction to obtain the flow intensity changes between functional cells in the future period, which provides quantitative support for traffic signal timing, travel route recommendation, congestion warning and regional resource scheduling. The core of the intelligent traffic prediction and control method based on dynamic functional cell coupling described in this application is: first, through the organic combination of motif analysis and graph neural network, to extract the high-level structural features of the network and generate the spatiotemporal embedding of nodes, so as to accurately reflect the changes in the strength of traffic connections within and outside the region during the dynamic division process, and use the weighted clustering algorithm to abstract homogeneous regions into functional cells, and then construct a mutual feedback time series sequence based on the historical flow data between functional cells and use LSTM to perform time series modeling and prediction on the sequence, so as to achieve accurate prediction of the future traffic flow intensity between functional cells, and finally convert the prediction results into executable traffic management strategies such as signal timing optimization, path induction and congestion warning.

[0107] This application introduces motif-based high-order network features and spatiotemporal clustering of functional cells to compensate for the neglect of regional-level coupling relationships in traditional traffic forecasting. By organically integrating dynamic graph neural networks and LSTM, it improves the perception of traffic network structure evolution and temporal changes, so that in complex scenarios such as morning and evening rush hours and holiday flow peaks, it can more accurately capture the nonlinear mutual feedback effects between regions and generate real-time traffic control strategies based on this, thereby significantly improving the efficiency of signal timing, optimizing path navigation effects, improving the accuracy of congestion warnings, and enhancing the system's adaptive response capabilities to emergencies and network structure changes.

[0108] As a preferred example of the present application, in step S1, the construction of the initial static city network includes the following specific steps: collecting the trajectory data of taxis, shared cars, etc., and taking the taxi area as the main geographical unit as the node set V0 in the city network, using the driving trajectories of all vehicles as the supplementary directed edge set E0 in the city network, and focusing on constructing the driving trajectories between regions, thereby forming a directed graph containing node sets and edge sets, that is, the static city network G0 =<V0,E0> . In the example of this application, the construction of the initial static urban network adopts a combination of multi-source data fusion and regional abstraction. First, the trajectory data of various types of urban vehicles including taxis and shared cars are collected through the vehicle positioning system, urban traffic management platform or third-party travel platform. These data contain elements such as time, location, speed and driving path of each vehicle, with high timeliness and geographical accuracy. Then, the city is divided according to the taxi operation area, and these areas are abstracted as nodes in the network. Each node corresponds to a clear geographical location and represents a type of traffic functional area, forming a node set V0 in the urban network. Then, the trajectory paths of all vehicles are analyzed, and the directed connection relationship between the nodes is determined based on the start and end areas traversed by the trajectory. The number of flows between each pair of areas in a specific time period is counted to construct a directed edge set E0, and each edge is given a weight representing the flow intensity. Finally, the directed graph G0 is formed by integration.<V0,E0> As a static urban network model, it provides structural support for subsequent dynamic network generation, traffic cell division and time series modeling.

[0109] This application uses trajectory data of multiple types of urban vehicles as the input information source and combines it with regional division to construct an urban traffic network, effectively improving the integrity and diversity of traffic network modeling and avoiding coverage blind spots or local deviations that may be caused by a single data source. At the same time, the directed edge sets constructed using trajectory flow relationships can more realistically reflect the directionality and intensity characteristics of traffic flow, thereby improving the static network's restoration of the city's real traffic conditions.

[0110] As a preferred example of the present application, in step S2, a trajectory containing information such as location and time is used as input. A fixed time interval Δ is given and divided into a discrete time interval set {Δ, 2Δ, ..., tΔ} based on 24 hours a day. Each time period represents an independent urban traffic snapshot. By summarizing and processing the vehicle trajectory data collected within each time period, the traffic interaction between different geographical locations in the city is extracted, and these geographical locations are abstracted into network nodes. The network nodes represent various geographical locations in the city, such as traffic intersections, commercial areas, residential areas, etc. The flow paths between regions are modeled as directed edges, and the directed edges represent the interaction relationships between these geographical locations, such as vehicle flow, personnel exchanges, etc., thereby forming a dynamic urban network G = {G0, G1, ..., G t The static urban network G at the tth time interval t = <V t ,E t >, where V t , E t Represents the city network node set and edge set at the tth time interval, each node Represents various geographical locations in the city, each edge It represents the interaction relationship between geographical locations (if the cells are reachable, there is an edge; if they are not reachable, there is no edge; i and j are two reachable cell nodes).

[0111] This application discretizes the city's daily time series into multiple time periods and constructs static urban network snapshots for each time period, achieving a fine-grained characterization of the city's traffic status. It can more accurately capture the time-varying characteristics of urban traffic conditions and provide more reliable data support for subsequent traffic forecasting and planning.

[0112] As a preferred example of this application, a static city network is composed of multiple city function cells. Each city functional cell All include G t A set of nodes and edges in the cell body, and the interaction strength inside the cell body is relatively strong, while the interaction strength outside the cell body (i.e. between the cell bodies) is relatively weak. t Each edge in Set a weight represents the node in the city network at the tth time interval and The interaction strength between them is normalized to limit the edge weights to the range of [0,1], and the urban functional cells are dynamically evolved with time intervals. In the process of building a dynamic urban network, in order to accurately model the traffic connection strength between urban functional cells, the system sets weights for all edges in the urban network snapshot generated at each time interval. For example, at the tth time interval, for each connecting node and edge The traffic interaction frequency between the two is counted based on the vehicle trajectory data, and its value is assigned to the edge as the initial interaction intensity value Since the nodes, edges and weights of dynamic urban networks change over time, the distribution of interaction intensity between different time periods and different regions may vary by orders of magnitude. Traditional methods often suffer from inconsistencies in data formats or differences in weight ranges, leading to low processing efficiency or unstable results. To solve this problem, this method normalizes all edge weights in each urban network snapshot. Mapping to the closed interval [0,1] makes the edge weights between different snapshots have a unified scale and are comparable, eliminating errors or instability caused by the uneven distribution of original data, thereby improving the adaptability and convergence of subsequent model processing.

[0113] As a preferred example of the present application, when performing the division of urban functional cells in step S3, the following steps are included:

[0114] S31: For a directed weighted dynamic city network G = {G0, G1, ..., G t}, based on the specified directed motif type, extract the motif features at each time step t and generate weighted motif strength and the motif matrix W M , and integrate historical information for time smoothing;

[0115] S32: Use graph neural network to aggregate node features and combine with directed weighted adjacency matrix A t and smooth motif matrix Generate node embedding Z t , and temporal smoothing is used to make the embedding continuous;

[0116] S33: Node-based embedding Z t , using weighted K-means algorithm to divide dynamic urban functional cells The cluster is initialized with the cell body centroid of the previous time step.

[0117] In the example of this application, during the process of urban functional cell division, the representative directed motif structure in the network is first extracted, and the weighted motif strength of the node is calculated by statistically multiplying the edge weights of the node in the motif. To construct the motif matrix W M In order to ensure the robustness of the feature, the motif strength Perform time smoothing, integrate historical information to reduce the interference caused by dynamic changes, and then use the powerful feature aggregation ability of graph neural network to transform motif intensity and the directed weighted adjacency matrix A t Combining and inputting two layers of graph neural network for feature aggregation, a richer and more accurate node embedding Z is generated. t These embeddings not only contain the local connection information of the nodes, but also incorporate the global motif structure information. Finally, a weighted K-means algorithm is used in this embedding space and the cluster center is initialized with the cell mass center of the previous time step, thereby dividing the dynamic urban functional cells to achieve accurate capture and stable tracking of the urban network structure. This application extracts motif features and smoothes them over time, uses graph neural networks to aggregate node features to generate embeddings, and then uses a weighted K-means algorithm based on the embeddings to divide the dynamic urban functional cells, achieving in-depth analysis and dynamic tracking of the urban network structure.

[0118] Traditional urban network analysis is mostly based on edge connections, ignoring high-order structures (such as loops or feedback patterns), resulting in insufficient feature expression, especially in dynamic urban networks. In addition, mutations in dynamic urban networks (such as sudden changes in nodes or edges) often cause feature instability. This application captures high-order topology and enhances continuity by performing motif feature extraction and time smoothing on each static urban network, that is, a snapshot of the city within each time period. Motif features enrich the structural representation of the network, and time smoothing reduces the interference of dynamic changes, improves the robustness of features and the accuracy of cell division, and provides a reliable basis for interaction prediction.

[0119] As a preferred example of this application, in the static urban network of each time period, in order to fully explore the high-order interaction characteristics in the transportation network and improve the expression ability of the node structure role, this method selects the directed triangle motif with cyclic representation meaning as the structural unit. In the motif feature extraction stage, the system first identifies all directed triangle motif instances involved by each node based on the graph structure, and multiplies the edge weights involved in each motif to obtain a weighted motif strength value reflecting the local interaction strength. These strengths are then aggregated per node to construct a weighted motif matrix to represent the high-order structural characteristics of the entire network. Considering that nodes and edges in dynamic networks may mutate in different time periods, resulting in feature instability, this method performs temporal smoothing on the motif strength value, that is, by setting a smoothing coefficient, the features of the previous time period are introduced into the current features to improve continuity, and the motif matrix is further normalized by column to unify the feature dimension and scale. The final smoothed and normalized motif matrix is used for graph neural network feature aggregation input, thereby providing a more stable and expressive structural input for subsequent node embedding and functional cell body partitioning.

[0120] In step S31, motif feature extraction is specifically as follows:

[0121] Motif selection: A directed triangle motif was chosen to reflect the circular interaction pattern in the urban network.

[0122] Motif strength calculation: calculation Weighted motif strength:

[0123]

[0124] in, Contains nodes The triangle set, the weight multiplication reflects the interaction strength.

[0125] Temporal smoothing: To avoid sudden changes, smooth the motif intensity:

[0126]

[0127] Where α is the first smoothing factor, such as α = 0.2, the initial

[0128] Construct a weighted motif matrix:

[0129]

[0130] Normalization:

[0131]

[0132] smooth:

[0133]

[0134] γ is the second smoothing factor.

[0135] In order to improve the efficiency of the algorithm, this step uses random walk sampling to calculate the motif, and the obtained and For use in the next step of GNN.

[0136] This application breaks through the limitations of structural expression based solely on adjacency relationships in traditional urban network analysis by extracting motif structural features that reflect cyclic interactions and performing temporal smoothing and normalization processing on them. It greatly enhances the ability of features to express the complex structural roles of nodes, and can maintain the continuity and robustness of feature expression in the face of node or edge mutations in dynamic networks, thereby effectively improving the stability of subsequent GNN aggregation node embedding, and also improving the accuracy and temporal consistency of urban functional cell division, providing a high-quality structural input basis for capturing regional linkage patterns in traffic forecasting, and helping to improve the overall modeling capability and prediction level of intelligent transportation systems for complex time-varying structures.

[0137] In traffic prediction in the prior art, traditional embedding methods ignore edge weights and directionality, and are sensitive to changes in dynamic networks, resulting in unstable embedding or insufficient expressiveness. When performing node feature aggregation in a dynamic urban network, the present application uses a graph neural network (GNN) to capture complex structures through a multi-layer aggregation strategy, combining motif features and time smoothing to adapt to dynamic urban networks. In step S32, by designing a motif-strong GNN and a time smoothing mechanism, node embedding integrates motif strength, degree information, and network topology, providing a robust feature representation, and time smoothing ensures the continuity of the dynamic network, improving the stability of cell division and prediction.

[0138] This application uses GNN aggregation node features when aggregating GNN features, combined with the directed weighted adjacency matrix A and the weighted motif matrix W M Generate node embeddings.

[0139] Initial features: calculated for each node Weighted motif strength: for each node Constructing feature vectors Including out-degree, in-degree, and smooth motif strength

[0140] Use two layers of GNN aggregation:

[0141]

[0142] in I is the identity matrix, which retains the node's own information by adding self-loops; H (l) is the node embedding of layer l, is the learnable weight, σ is the activation function; yes The out-degree matrix is used to fuse the edge weight information and motif information; λ∈[0,1] is a hyperparameter, such as λ=0.3 to balance the directed weighted adjacency matrix A and the smoothed motif matrix.

[0143] Embed the third smoothing factor β: β = 0.2, on this basis, the node embedding Z is obtained t , the formula is as follows:

[0144]

[0145] When performing node feature aggregation in a dynamic urban network, this application adopts a two-layer aggregation strategy in the graph neural network (GNN) in order to more comprehensively integrate structural information and interaction weights. In the first layer aggregation stage, the unit matrix is added to the directed adjacency matrix A to enhance the node's own feature expression. At the same time, the node out-degree matrix is combined to form a combined matrix for integrating edge weight information and smoothing motif matrix information. By setting an adjustable hyperparameter λ, the influence of these two types of structural inputs is dynamically balanced, making the model more adaptable in different traffic structure environments. In the second layer aggregation, the node embedding Z is further refined. t In order to improve its representation ability and enhance the nonlinear expression depth of the model, in order to solve the problem of mutations that may occur in node embeddings between different time steps, a time smoothing mechanism is introduced in the method. The node embedding generated at the current moment is weighted and fused with the result of the previous time step through a preset smoothing factor β, so that the embedding result still has smooth continuity under dynamic changes, thereby providing a more stable embedding foundation for the continuous division of functional cell bodies and interaction modeling.

[0146] The cell structure of dynamic urban networks evolves over time. Traditional partitioning methods (such as Louvain or spectral clustering) lack support for weights and directionality and are prone to unstable or fragmented cells in dynamic networks. This application uses weighted K-means and dynamic centroid initialization to generate continuous and meaningful cell partitions when partitioning dynamic urban functional cells. This simplifies complex networks into modular structures, facilitating the analysis of inter-regional interactions. Dynamic adjustments ensure that the partitions adapt to changes in the urban network, improving the accuracy and stability of predictions.

[0147] As a preferred example of this application, in step S33, the geometric distance and node weight of the embedding space are used to adopt the weighted K-means method to identify dynamic urban function cells. When initializing the cluster center, the centroid of the urban function cell at t-1 is referenced. The optimization goal is as follows:

[0148]

[0149] in, is a node The weight, combining the smoothing motif strength and degree, μ t,i It is the functional cell of the city The weighted centroid of is as follows:

[0150]

[0151] When t=0, randomly initialize k0 (10, which can be set according to the actual city situation) centroids. If t>=1, use The center of mass μ t-1,i As the initial value, add or delete the centroid to match k t .

[0152] Adjustment of the number of functional cell bodies:

[0153]

[0154] Where δ is a preset adjustment parameter, which is dynamically adjusted according to the network scale, such as δ = 1.0.

[0155] In the process of dynamic urban functional cell identification and adjustment, in order to improve the temporal consistency and structural rationality of the division results, this application first constructs a comprehensive weight for each node by combining the smoothed motif strength, out-degree and in-degree information after the node embedding is generated. This weight not only reflects the structural centrality of the node, but also reflects the degree of its interaction in the motif. Subsequently, the weighted K-means algorithm is used in the embedding space to divide the urban functional cell. At t=0, the system randomly initializes several centroids. In subsequent time steps of t≥1, the centroid of the urban functional cell identified in the previous time step is preferentially inherited as the initial clustering center, and the centroid is automatically added or deleted according to the current network structure changes to dynamically adapt to actual changes. When dividing the functional cells of dynamic cities, this application constructs a weighted similarity index by fusing multi-source structural features, which significantly improves the accuracy and robustness of the functional cell division. In particular, in the face of scenarios where node connections change frequently and traffic flows fluctuate violently in dynamic urban networks, the weighted K-means clustering based on the dual information of node embedding and structural weight can more effectively reflect the structural consistency within the urban area. At the same time, cluster initialization and functional cell state change identification based on historical centroids ensure the continuity and stability of the division results in the time dimension, thereby providing an analytical basis with a clear structure and controllable changes for subsequent traffic flow prediction, congestion analysis and regional scheduling strategies.

[0156] As a preferred example of the present application, when predicting the change in the mutual feedback strength between functional cell bodies in subsequent time segments, the following steps are included:

[0157] S41: Extracting cell-level features, including city function cell embedding Motif intensity and urban functional cell out / in degree are used to construct the mutual feedback characteristics of urban functional cell pairs. Based on the historical interaction feature sequence, the time series model is used to predict the basic mutual feedback strength of the urban functional cell pairs in the future time step t+h.

[0158] S42: Combined with the dynamic state of urban functional cells, including birth, expansion, shrinkage, and death, the interaction intensity prediction value is adjusted through weighted similarity matching

[0159] In the mutual feedback intensity prediction process, this application first constructs a multi-dimensional feature vector that integrates the cell embedding position, motif interaction intensity, and in-and-out activity to fully characterize the structural performance and connection activity of the urban functional cell body. The feature vectors are then combined in pairs to form mutual feedback features representing regional interaction relationships, and are organized into time series and input into the LSTM model to utilize the model to conduct deep learning on the potential temporal dependencies in the historical sequence, thereby predicting the mutual feedback intensity in future time steps. On this basis, the system detects the birth, expansion, shrinkage, or death status of the urban functional cell body in real time, and dynamically corrects the LSTM output by weighted similarity matching with the historical cell body, thereby forming a final mutual feedback prediction result consistent with the actual network evolution.

[0160] Among intelligent traffic prediction methods, traditional time series prediction approaches focus on a single time series, ignoring network structure and regional interactions, resulting in limited prediction accuracy, especially in dynamic urban networks. Cell-level interaction prediction requires the integration of topological and temporal information. This application achieves accurate prediction through cell-level embedding and LSTM models. Feedback strength prediction provides insight into the future state of the network, supporting the optimization and management of dynamic networks. This cell-level perspective simplifies complex network analysis, improving computational efficiency and forecast interpretability.

[0161] As a preferred example of the present application, step S41 includes:

[0162] S411: Cell level feature extraction:

[0163] In getting the node embedding Z t Based on the The embedding formula is as follows:

[0164]

[0165] S412: Construct the basic mutual feedback characteristics of the cell body;

[0166] First, for each cell body Constructing feature vectors The formula is as follows:

[0167]

[0168] in, is the sum of the motif intensities of all nodes inside the cell body, are the sum of the out-degree and in-degree of the nodes inside the cell body respectively;

[0169] Then, construct the mutual feedback feature between cell pairs The formula is as follows:

[0170]

[0171] in Indicates that at time t, the cell body The basic mutual feedback strength between is as follows:

[0172]

[0173] in, It's the edge weight at time t;

[0174] S413: Time Series Modeling;

[0175] Use LSTM to model the time series of the mutual feedback strength of urban cell pairs and input historical mutual feedback features The hidden layer dimension is 128, τ=5, and the LSTM model is constructed as follows:

[0176] h t ,c t =LSTM(X i,j ,h t-1 ,c t-1 )

[0177]

[0178] where h t ,c t is the hidden state and cell state of LSTM; W out is the weight of the output layer, b out is the output layer bias, and the loss function is:

[0179]

[0180] N is the number of city cell pairs, is the basic mutual feedback strength of the city functional cell pair at that moment, is the predicted value of interaction strength.

[0181] In the cell-level mutual feedback prediction process, this application first aggregates the motif strength, node out-degree, and in-degree information of all nodes within each functional cell body based on the node embedding obtained in the previous stage to construct a cell body embedding vector, where the motif strength is used to characterize the structural tightness between nodes within the cell body, and the node out-degree and in-degree reflect the output and input connection capabilities of the cell body, respectively. Then, the edge weights between any two cell bodies in the urban network graph are extracted as the basis for their interaction strength, and the mutual feedback feature vectors of the cell pairs are comprehensively constructed. The mutual feedback feature sequence between each pair of cell bodies is constructed as a time series input. Finally, these sequences are learned and modeled using an LSTM model. Through training, the inherent laws of the evolution of the mutual feedback features over time are captured, and the changes in the mutual feedback strength between each cell pair in the future are predicted, thereby realizing the time series perception and quantitative expression of regional-level traffic evolution trends, and further realizing forward-looking analysis and judgment of regional interaction patterns in dynamic traffic networks.

[0182] As a preferred example of this application, in step S42, by calculating the weighted similarity between dynamic urban function cells, and The cell state of the city cell body is:

[0183]

[0184] The birth of dynamic urban functional cells: With all of but It is the functional cell of the new city;

[0185] Dynamic urban functional cell death: if With all of but die;

[0186] Dynamic urban functional cell body continues: if and satisfy and but yes Continuation of

[0187] Dynamic urban functional cell expansion: and but merge

[0188] Dynamic urban functional cell shrinkage: if and The city's functional cell body is shrinking;

[0189] Dynamic city cell birth: if is a newly born cell body, and the predicted value of the mutual feedback strength of the new cell body is initialized based on the average mutual feedback strength of the current time step:

[0190]

[0191] Dynamic city cell death: if Died at time t+h, that is, there is no matching cell body,

[0192] Dynamic city cell expansion and contraction: If and Match, adjust for scale changes:

[0193]

[0194] in is the basic mutual feedback strength of the city functional cell pair at that moment,

[0195] is the predicted value of mutual feed strength.

[0196] The cell structure of a dynamic urban network changes over time (such as the addition of new nodes or the disappearance of cells). Traditional prediction methods ignore these evolutions, resulting in unreasonable or inaccurate prediction results. In the example of this application, in order to adapt to the changes brought about by the structural evolution of the urban traffic network and maintain the rationality and continuity of the mutual feedback prediction results, this application introduces a state recognition and mutual feedback intensity adjustment mechanism for dynamic urban functional cells in the prediction process. First, by comparing the structural characteristics and distribution of urban functional cells in consecutive time steps, the weighted similarity between each pair of cells is calculated. This similarity comprehensively considers indicators such as node composition overlap, node embedding feature distance, and overall scale similarity. If the similarity of a certain cell to all other cells in the current time step is lower than the set threshold, it is identified as a new urban functional cell; if it has no match in subsequent time steps, If the cell body is matched, it is judged to be in a dead state; if the similarity between the current cell body and the cell body at a previous time is high and the node scale is basically the same, it is judged to be a continuous state; if the similarity is high but the scale increases significantly, it is considered to be expanding, and if the number of nodes decreases, it is considered to be shrinking. After completing the state identification, the mutual feedback strength between the cell bodies is corrected and adjusted according to different states. The new cell body uses the mean value of the current overall mutual feedback strength to initialize the mutual feedback relationship. The mutual feedback strength of the dead cell body is reset to zero to reflect its exit from the network. The mutual feedback strength of the expansion or shrinkage state is linearly adjusted according to the cell body scale change ratio, thereby ensuring the reasonable transition of the network mutual feedback structure in time sequence.

[0197] By introducing a weighted similarity recognition mechanism and a mutual feedback strength state adjustment strategy, this application can still achieve accurate judgment and efficient response to the state of functional cells in the context of frequent changes in dynamic urban network structures, effectively avoiding the distortion of prediction results caused by the emergence of new areas or the disappearance of old areas. At the same time, by performing state matching adjustment on the mutual feedback strength, it ensures that the prediction results are highly consistent with the actual state of the network, significantly improving the adaptability and prediction stability of the intelligent transportation system to the evolution of urban transportation structure.

[0198] As a preferred example of the present application, the present application also discloses an intelligent traffic prediction control system based on dynamic functional cell coupling, comprising:

[0199] A data acquisition module is used to collect vehicle trajectory data and construct a city static traffic network based on taxi operating areas as basic geographical units. The nodes of the static traffic network represent geographical locations, and the edges represent traffic interactions between nodes and have weighted directions.

[0200] The dynamic network construction module is used to construct the static network into a sequence of urban network snapshots at multiple times according to a given time interval Δ, thereby forming a dynamic urban network;

[0201] A feature extraction and embedding module, comprising a motif extraction unit and a graph neural network (GNN) unit. The motif extraction unit is used to identify the directed motif structure in the traffic network at each moment and calculate the strength of the node's participation in the motif. The GNN unit is used to fuse the directed weighted adjacency matrix with the weighted motif matrix and generate node embeddings by aggregating node features.

[0202] The dynamic urban cell recognition module is used to divide urban functional cells using the weighted K-means method based on the node embedding results and combined with the time smoothing strategy, and to identify and update the cell state (birth, death, expansion, and contraction);

[0203] The time series prediction module extracts cell-level features and constructs the mutual feedback features of urban functional cell pairs. It is used to model the cell mutual feedback sequence based on the long short-term memory network (LSTM) and predict the traffic mutual feedback intensity between each cell pair in the future time period.

[0204] The strategy output module is used to generate traffic scheduling recommendations based on the prediction results to support traffic management tasks such as signal timing optimization, route navigation, congestion warning, and infrastructure planning.

[0205] The intelligent traffic prediction and control system based on dynamic functional cell coupling described in this application constructs static and dynamic traffic networks through multi-source trajectory data, inputs the motif reflecting the high-order structure and basic adjacency information into the graph neural network to generate node embedding, and then combines time smoothing and weighted clustering to dynamically divide the functional cells and identify their evolutionary states. By aggregating cell-level embeddings, motif strength and in-and-out degree, a mutual feedback feature sequence is constructed and time series prediction is performed by LSTM, and finally the predicted mutual feedback strength results are converted into traffic management strategies such as signal timing, navigation path and congestion warning. Specifically, the intelligent traffic prediction and control system based on dynamic functional cell coupling described in this application first collects high-precision trajectory data of taxis and shared vehicles with time and speed information through the urban traffic management platform and the third-party travel interface in the data acquisition stage, and abstracts the taxi service area as the geographical unit into a network node, converts the vehicle flow path between regions into weighted directed edges to construct a static traffic network, and then generates a series of network snapshots reflecting the urban traffic status at different time periods at a preset time interval (given time interval) in the dynamic network construction stage and forms a dynamic graph, and identifies the directed motif reflecting the regional cyclic interaction in each snapshot through the motif extraction unit and calculates the node The weighted motif strength of each point is used to construct a motif matrix. The GNN unit then fuses the adjacency matrix and the motif matrix for multi-layer aggregation to generate node embeddings. Then, in the dynamic urban cell recognition module, time-smoothed embedding is combined with node weights to dynamically divide functional cells using the weighted K-means algorithm and track their birth, expansion, shrinkage, and extinction states. In the time series prediction stage, the embedding, motif strength, and in-and-out degree of each functional cell are calculated for each pair. The inter-feedback feature sequence is input into the LSTM model to predict future inter-feedback strength. Finally, the strategy output module provides executable scheduling solutions for traffic light timing, route guidance, congestion warning, and regional planning based on the prediction results.

[0206] Those skilled in the art will understand that all or part of the steps for implementing the above-mentioned method embodiments can be accomplished by hardware related to the computer program. The aforementioned computer program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above-mentioned method embodiments; and the aforementioned storage medium includes: ROM, RAM, FLASH, floppy disk, magnetic disk or optical disk, mechanical hard disk, cloud and other media that can store program codes. The computer processor can be a central processing unit (CPU), a microprocessor (MPU), a digital signal processor (DSP), or a field programmable gate array (FPGA), that is, the steps of the intelligent traffic prediction control method based on dynamic functional cell coupling can be implemented by the above-mentioned processor.

[0207] The embodiments of the present application are described above in conjunction with the accompanying drawings. Unless there is a conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of this application, ordinary technicians in this field can also make many forms without departing from the purpose of this application and the scope of protection of the claims, all of which are within the protection of this application.

Claims

1. An intelligent traffic prediction control method based on dynamic functional cell coupling, characterized in that: include: Obtain urban vehicle trajectory data and construct an initial static urban network using taxi areas as geographical units. Network nodes represent the geographical locations of the city, and network edges represent the traffic interaction relationships between nodes with directions and weights. The static urban network is transformed into a dynamic urban network consisting of multiple time-segment urban network snapshots to characterize the time-varying structure of the urban transportation network; In each time segment network snapshot, motif features are extracted and a feature matrix is constructed. The GNN model is used to aggregate the dynamic embedding of geographic locations to divide urban functional cells, and the dynamic changes of cell bodies are identified in combination with the cell body state at the previous moment. Based on the dynamic urban functional cell structure, the mutual feedback feature vectors between functional cells are constructed and the mutual feedback time series is formed. The LSTM model is used to predict the changes in the mutual feedback intensity between functional cells in subsequent time segments. Provide support for urban traffic planning based on the prediction results, including at least one of signal light timing, route navigation, traffic warning and regional scheduling suggestions.

2. The intelligent traffic prediction control method based on dynamic functional cell coupling according to claim 1 is characterized in that: In the process of constructing a dynamic urban network, according to a given time interval Δ, 24 hours a day is divided into a discrete time interval set {Δ, 2Δ, ..., tΔ}, and a city network snapshot is generated for each time interval. The dynamic urban network is represented as a set of city snapshots at multiple time intervals G = {G0, G1, ..., G t }, where G t = <V t ,E t > represents the static urban network at the tth time interval, V t 、E t They represent the city network node set and directed edge set at the tth time interval, G t Each edge in Set a weight represents the node in the city network at the tth time interval and The interaction strength between them is normalized to limit the edge weight to the range of [0,1].

3. The intelligent traffic prediction control method based on dynamic functional cell coupling according to claim 2 is characterized in that: When dividing the urban functional cells, the following steps are included: S31: For a directed weighted dynamic city network G = {G0, G1, ..., G t }, based on the specified directed motif type, extract the motif features at each time step t and generate weighted motif strength and the motif matrix W M , and integrate historical information for time smoothing; S32: Directed weighted adjacency matrix A t and smooth motif matrix Input graph neural network for feature aggregation and generate node embedding Z t , and temporal smoothing is used to make the embedding continuous; S33: Node-based embedding Z t , using weighted K-means algorithm to divide dynamic urban functional cells The cluster is initialized with the cell body centroid of the previous time step.

4. The intelligent traffic prediction control method based on dynamic functional cell coupling according to claim 3 is characterized in that: In step S31, the motif feature extraction includes: S311: Compute node The number of motifs at time t and weighted motif strength The formula is as follows: Where A is G t The directed weighted adjacency matrix of Contains nodes A collection of triangles, is a node The product of the weights of all edges in M, where M is the participating motif instance; S312: Weighted motif intensity using temporal smoothing Get smooth motif intensity Where α is the first smoothing factor, the initial S313: Construct weighted motif matrix W M , Representation node The weighted connection strength in a directed motif is given by: Normalized and smoothed to obtain a smooth motif matrix: γ is the second smoothing factor.

5. The intelligent traffic prediction control method based on dynamic functional cell coupling according to claim 4 is characterized in that: In step S32, the aggregation formula of GNN is: in I is the identity matrix, H (l) is the node embedding of layer l, is the learnable weight, σ is the activation function, yes The out-degree matrix is used to fuse the edge weight information and motif information, λ∈[0,1], and the node embedding Z is obtained on this basis. t , the formula is as follows: Where β is the third smoothing factor.

6. The intelligent traffic prediction control method based on dynamic functional cell coupling according to claim 5 is characterized in that: In step S33, the weighted K-means optimization objective is: in, is a node The weight, combining motif strength and degree, μ t,i It is the functional cell of the city The weighted centroid of t is the number of urban functional cells.

7. The intelligent traffic prediction control method based on dynamic functional cell coupling according to any one of claims 1 to 6, characterized in that: When predicting the change in the mutual feedback strength between functional cell bodies in subsequent time segments, the following steps are included: S41: Extracting cell-level features, including city function cell embedding Motif intensity and urban functional cell out / in degree are used to construct the mutual feedback characteristics of urban functional cell pairs. Based on the historical interaction feature sequence, the time series model is used to predict the basic mutual feedback strength of the urban functional cell pairs in the future time step t+h. S42: Combined with the dynamic state of urban functional cells, including birth, expansion, shrinkage, and death, the interaction intensity prediction value is adjusted through weighted similarity matching 8. The intelligent traffic prediction control method based on dynamic functional cell coupling according to claim 7 is characterized in that: In step S41, it includes: S411: Cell level feature extraction: In getting the node embedding Z t Based on the The embedding formula is as follows: S412: Construct the basic mutual feedback characteristics of the cell body; First, for each cell body Constructing feature vectors The formula is as follows: in, is the sum of the motif intensities of all nodes inside the cell body, are the sum of the out-degree and in-degree of the nodes inside the cell body respectively; Then, construct the mutual feedback feature between cell pairs The formula is as follows: in Indicates that at time t, the cell body The basic mutual feedback strength between is as follows: in, It's the edge The weight at time t; S413: Time Series Modeling; Use LSTM to model the time series of the mutual feedback strength of urban cell pairs and input historical mutual feedback features The LSTM model is constructed as follows: h t ,c t =LSTM(X i,j ,h t-1 ,c t-1 ) where h t ,c t is the hidden state and cell state of LSTM; W out is the weight of the output layer, b out is the output layer bias, and the loss function is: Where N is the number of city cell pairs, is the basic mutual feedback strength of the urban functional cell pair at time t+h, is the predicted value of LSTM mutual feedback strength at time t+h.

9. The intelligent traffic prediction control method based on dynamic functional cell coupling according to claim 8 is characterized in that: In step S42, the weighted similarity between dynamic urban function cells is calculated to identify and The cell state of the city cell body is: like With all of but Initialize the mutual feedback strength of the new urban functional cell: like With all of but die, like and satisfy and but yes Continuation of like and but merge The city's functional cell body expands; like and The city's functional cell body is shrinking; when and Match, adjust for scale changes: in is the basic mutual feedback strength of the city functional cell pair at that moment, is the predicted value of mutual feed strength.

10. An intelligent traffic prediction control system based on dynamic functional cell coupling, characterized in that: include: Data acquisition module, used to collect vehicle trajectory data and build a static urban traffic network; Dynamic network construction module, used to construct a static network into a sequence of city network snapshots at multiple times to form a dynamic city network; Feature extraction and embedding module, including a motif extraction unit for identifying directed motif structures and calculating node participation strength, and a graph neural network (GNN) unit for fusing the adjacency matrix with the motif matrix and generating node embeddings; The dynamic urban cell recognition module is used to divide urban functional cells using the weighted K-means method based on the node embedding results and combined with the time smoothing strategy, and to identify and update the cell state; The time series prediction module extracts cell-level features and constructs the mutual feedback features of urban functional cell pairs. This module is used to model the cell mutual feedback sequence based on the long short-term memory network and predict the traffic mutual feedback intensity between each cell pair in the future time period. The strategy output module is used to generate traffic scheduling recommendations based on the prediction results to support any traffic management task such as signal timing optimization, route navigation, congestion warning, and infrastructure planning.

Images