A public transportation scheduling optimization method driven by multimodal time series data analysis

Abstract

The present invention relates to the field of intelligent traffic management technology and discloses a public transportation scheduling optimization method driven by multimodal time series data analysis, comprising: obtaining multimodal traffic data and performing spatiotemporal alignment preprocessing to extract feature vectors; fusing multiple feature vectors based on a cross-modal interactive learning method; performing real-time prediction of public transportation conditions based on the multimodal fused feature vectors to obtain a public transportation flow prediction value, and formulating a public transportation scheduling strategy based on the obtained public transportation flow prediction value. The present invention fuses multiple feature vectors based on a cross-modal interactive learning method, effectively capturing the interrelationships between different types of data, and formulating a public transportation scheduling strategy based on a particle swarm optimization algorithm, effectively balancing the capabilities of global and local searches. This can effectively improve the performance of the PSO algorithm in public transportation vehicle scheduling problems and achieve a more efficient scheduling solution.

Description

Technical Field

[0001] The present invention relates to the field of intelligent traffic management technology, and more particularly to a public transportation scheduling optimization method driven by multimodal time series data analysis. Background Art

[0002] With the acceleration of urbanization, public transportation systems face unprecedented challenges. On the one hand, the growth and expansion of urban populations has led to a sharp increase in demand for public transportation. On the other hand, increasingly serious problems such as traffic congestion and environmental pollution have made improving the efficiency and service quality of public transportation systems a key task in urban development. Traditional public transportation scheduling relies primarily on empirical rules and historical data. This approach is inadequate for dynamically changing traffic conditions and struggles to meet the requirements of real-time and accuracy.

[0003] Against this backdrop, optimizing public transportation scheduling using advanced data analytics techniques has become a research hotspot. In particular, multimodal time-series data analysis, as an emerging technology, can effectively integrate data from diverse sources. However, effectively fusing these diverse data types and extracting useful information to support decision-making remains a complex problem.

[0004] Therefore, how to provide a public transportation scheduling optimization method driven by multimodal time series data analysis is an urgent problem that needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a public transportation scheduling optimization method driven by multimodal time series data analysis, which can not only improve the response speed and accuracy of the public transportation system, but also enhance its ability to adapt to dynamic environments, thereby providing passengers with more efficient and convenient services.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A public transportation scheduling optimization method driven by multimodal time series data analysis, comprising:

[0008] Acquire multimodal traffic data and perform spatiotemporal alignment preprocessing to extract feature vectors;

[0009] Fusion of multiple feature vectors based on cross-modal interactive learning method;

[0010] Real-time prediction of public transportation conditions is performed based on multimodal fusion feature vectors to obtain public transportation flow prediction values;

[0011] Formulate public transportation scheduling strategies based on public transportation flow forecast values.

[0012] Preferably, the multimodal traffic data includes vehicle-mounted satellite positioning data, traffic flow data, and real-time data fed back by social software, wherein the structured data is traffic flow data, the semi-structured data includes vehicle-mounted satellite positioning data, and the unstructured data includes text data fed back by social software.

[0013] Preferably, extracting the feature vector includes:

[0014] Processing structured data based on autoencoders;

[0015] Construct an urban traffic network map and extract spatial features from semi-structured data based on a graph attention network. The adjacency matrix of the graph attention network is constructed based on the road topology.

[0016] Time series analysis and key information extraction of unstructured data based on bidirectional long short-term memory network.

[0017] Preferably, multiple feature vectors are fused based on a cross-modal interactive learning method, including:

[0018] (1) A multimodal spatiotemporal correlation graph is constructed based on the urban traffic network graph. The nodes represent the feature vectors of different modalities, and the edge weights are calculated through the spatiotemporal attention mechanism. The edge weight calculation process is:

[0019] Calculate modality attention weights:

[0020]

[0021] in, They represent the query transformation matrix and the key transformation matrix respectively, K represents the number of attention heads, and D is the feature dimension after transformation. and Represent the feature vectors of node i and node j respectively, is the modality attention weight;

[0022] Calculate the attention weight based on spatial distance calculation:

[0023]

[0024] in, Represents the Euclidean distance of the grid center point, is the time attenuation coefficient, Represents the attention weight calculated based on spatial distance;

[0025] Calculate the attention weight based on the time difference:

[0026]

[0027] in, represents the time decay coefficient, Represents the time interval between two nodes, Represents the attention weight calculated based on the time difference;

[0028] Normalized attention weights:

[0029]

[0030] in, represents the activation function, represents the final attention weight from node i to node j;

[0031] (2) Aggregating multimodal features through graph attention networks, including:

[0032] Use K independent attention heads to parallelize the computation:

[0033]

[0034] in, , represents the value transformation matrix of the k-th attention head, represents the neighbor nodes of node i, represents the normalized attention weight of the k-th attention head, Represents the features of the kth attention head;

[0035] After parallel calculation of multiple attention heads, multi-head feature splicing is performed to output intermediate fusion features;

[0036] Perform residual connection and normalization on the intermediate fusion features;

[0037] (3) After the graph attention network aggregates the multimodal features, the node features are output through multi-layer spatiotemporal graph convolution;

[0038] The final node features are mapped into multimodal fusion feature vectors through the fully connected layer.

[0039] Preferably, the real-time prediction of public transportation conditions based on the multimodal fusion feature vector includes:

[0040] Use the spatiotemporal graph convolutional network to jointly extract spatiotemporal features from the fusion features;

[0041] A multi-scale time window mechanism is used to capture temporal dependencies of different granularities;

[0042] The feature weights are dynamically adjusted through the spatiotemporal gating unit to output the public transportation flow prediction value.

[0043] Preferably, a public transportation scheduling strategy is formulated based on a particle swarm optimization algorithm, including:

[0044] Use the multimodal fusion feature vector as the initial position and define the adaptability function;

[0045] Calculate particle fitness based on public transportation flow prediction values;

[0046] Update the position and velocity of each particle according to its fitness, taking into account the spatiotemporal constraints during the update process;

[0047] The update process is terminated when it reaches the maximum number of iterations or the fitness change is less than the set threshold, and the global optimal solution is obtained. The global optimal solution is the public transportation scheduling strategy.

[0048] Preferably, the position and velocity update formula of each particle is:

[0049]

[0050] in, is the updated speed, is the velocity of particle n in dimension m, is the position of particle n in dimension m, is the shrinkage factor, and are the maximum and minimum values ​​of the inertia weight, t is the current number of iterations, is the maximum number of iterations, are the initial and final learning factors respectively The value of are the initial and final learning factors respectively The value of is the historical optimal position of particle n in the mth dimension, is the optimal position of the entire particle swarm in the mth dimension, and is a random number between [0, 1];

[0051]

[0052] in, is the updated location.

[0053] Preferably, the public transportation scheduling strategy includes:

[0054] Vehicle departure time: the time when each vehicle departs from the departure station;

[0055] Number of vehicles: the number of vehicles required on a certain route;

[0056] Frequency of vehicle operation: the interval between vehicles;

[0057] Vehicle operation path: whether the vehicle needs to detour certain sections of road or adjust the operation route;

[0058] Stop time at each station: the time the vehicle stays at each station.

[0059] A computer device, characterized in that it includes: a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, a public transportation scheduling optimization method driven by multimodal time series data analysis is implemented.

[0060] A computer-readable storage medium is characterized in that a computer program is stored on the storage medium, and when the computer program is executed by a processor, a public transportation scheduling optimization method driven by multimodal time series data analysis is implemented.

[0061] As can be seen from the above technical solutions, compared with the existing technology, the present invention provides a public transportation scheduling optimization method driven by multimodal time series data analysis. It fuses multiple feature vectors based on a cross-modal interactive learning method, effectively capturing the relationships between different types of data. It also formulates a public transportation scheduling strategy based on a particle swarm optimization algorithm, effectively balancing the capabilities of global and local searches to achieve more efficient scheduling solutions. The overall technical framework of "spatiotemporal alignment - spatiotemporal feature extraction - spatiotemporal attention fusion - spatiotemporal joint prediction - constrained optimization scheduling" is formed. Therefore, the present invention not only achieves a comprehensive perception and understanding of the urban public transportation system, but also effectively improves the scientific nature and rationality of scheduling decisions, ultimately achieving the goal of improving transportation efficiency and service quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0063] Figure 1 This is a flow chart of a public transportation scheduling optimization method driven by multimodal time series data analysis provided by the present invention.

[0064] Figure 2 The present invention provides a flowchart for fusing multiple feature vectors based on a cross-modal interactive learning method.

[0065] Figure 3 This is a flow chart of formulating a public transportation scheduling strategy based on the particle swarm optimization algorithm provided by the present invention. DETAILED DESCRIPTION

[0066] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0067] The embodiment of the present invention discloses a public transportation scheduling optimization method driven by multimodal time series data analysis, such as Figure 1 Shown, including:

[0068] Acquire multimodal traffic data and perform spatiotemporal alignment preprocessing to extract feature vectors. Modal traffic data includes vehicle-mounted satellite positioning data, traffic flow data, and real-time data fed back by social software. In addition to the above data, other modal traffic data can also be obtained according to actual needs. Structured data is traffic flow data, including road section ID, statistical time window (such as every 5 minutes), number of vehicles, average speed, congestion level, etc. Semi-structured data includes vehicle-mounted satellite positioning data. Satellite positioning data contains vehicle ID, latitude and longitude coordinates, timestamp, instantaneous speed, direction and other information. Unstructured data includes text data fed back by social software, including but not limited to text content such as traffic accidents or emergencies on the road, large-scale event announcements, release time, and geographic location tags shared through social software; spatiotemporal alignment preprocessing includes a time alignment module and a space alignment module. The time alignment module uses a dynamic time warping algorithm to align the timestamps of multimodal data, and the space alignment module unifies the spatial coordinate system through geographic grid encoding;

[0069] Time-aligned asynchronous data:

[0070] Use traffic flow data as the benchmark time axis (e.g. fixed 5-minute intervals);

[0071] The vehicle-borne satellite positioning data is linearly interpolated to generate a coordinate sequence aligned with the reference time axis;

[0072] Social media texts are aggregated by time window, the texts are assigned to the nearest benchmark time window according to the publishing time, and event features are extracted.

[0073] Spatial alignment:

[0074] Define a city geographic grid (e.g. 500m x 500m grid):

[0075] In the vehicle-mounted satellite positioning data, the latitude and longitude of each trajectory point are converted into a grid ID, and the number of vehicles and average speed in each grid are counted.

[0076] In the traffic flow data, the road segment ID is mapped to the set of covered grids (e.g., road segment A spans grids G0235 and G0236), and the flow value is allocated according to the length ratio.

[0077] We perform location entity recognition on social text data. This involves using a geographic named entity recognition (GeoNER) model to extract locations (such as "Intersection A") from text and associate them with grid IDs. For text without explicit location descriptions, we map them based on the user's IP address or posted location tags.

[0078] The present invention solves the problems of temporal asynchrony and spatial heterogeneity through dynamic time warping and geographic grid coding.

[0079] Fusion of multiple feature vectors based on cross-modal interactive learning method;

[0080] Real-time prediction of public transportation conditions is performed based on multimodal fusion feature vectors to obtain public transportation flow prediction values;

[0081] Formulate public transportation scheduling strategies based on public transportation flow forecast values.

[0082] In this embodiment, extracting the feature vector includes:

[0083] The autoencoder extracts feature vectors from the preprocessed structured data. Specifically, the autoencoder learns a low-dimensional representation by encoding and decoding the structured data. The feature vector has the characteristics of noise reduction and can better reflect the true distribution of the data.

[0084] Construct an urban traffic network map. Use the graph attention network to extract spatial features from semi-structured data and obtain feature vectors. The adjacency matrix of the graph attention network is constructed based on the road topology. The specific process of constructing the urban traffic network map is as follows:

[0085] Obtain urban road network data, usually from open mapping services (such as OpenStreetMap) or GIS data from traffic management departments.

[0086] The key points of intersections or road sections are used as nodes in the graph, and each node contains longitude and latitude coordinates. For example, the start / end point of each intersection or road section is set as a node.

[0087] Edges are defined based on the actual connectivity of the roads. For example, if two roads are connected at an intersection, an edge is created between the corresponding nodes.

[0088] Add attributes to nodes and edges, such as node coordinates, road type (main road, secondary road), number of lanes, speed limit, etc.

[0089] The road sections and intersections of the urban traffic network map will be mapped to the corresponding geographic grid.

[0090] Based on the bidirectional long short-term memory network, time series analysis and key information extraction are performed on the pre-processed unstructured data, and the specific processing process is also achieved through existing technologies.

[0091] In this embodiment, multiple feature vectors are fused based on a cross-modal interactive learning method. In the cross-modal interactive learning method, fusing multiple feature vectors to generate a richer and more comprehensive multimodal representation is a core step. Figure 2 As shown, specifically including:

[0092] Each space-time node Corresponding to a geographic grid i and time window t, the nodes represent different modal feature vectors, and the edge weights are calculated through the spatiotemporal attention mechanism. The calculation formula is:

[0093]

[0094] in, They represent the query transformation matrix and the key transformation matrix respectively, K represents the number of attention heads, and D is the feature dimension after transformation. and Represent the feature vectors of node i and node j respectively, is the modality attention weight;

[0095]

[0096] in, Represents the Euclidean distance of the grid center point, is the time attenuation coefficient, Represents the attention weight calculated based on spatial distance;

[0097]

[0098] in, represents the time decay coefficient, Represents the time interval between two nodes, Represents the attention weight calculated based on the time difference;

[0099]

[0100] in, represents the activation function, represents the final attention weight from node i to node j;

[0101] Aggregate multimodal features through graph attention network, including:

[0102] Use K independent attention heads to parallelize the computation:

[0103]

[0104] in, , represents the value transformation matrix of the k-th attention head, represents the neighbor nodes of node i, represents the normalized attention weight of the k-th attention head, Represents the features of the kth attention head;

[0105] After parallel calculation of multiple attention heads, multi-head feature splicing is performed to output intermediate fusion features;

[0106] Perform residual connection and normalization on the intermediate fusion features;

[0107] After the multimodal features are aggregated by the Graph Attention Network (GAT), node features are output through multi-layer spatiotemporal graph convolution;

[0108] The final node features are mapped into multimodal fusion feature vectors through the fully connected layer.

[0109] Cross-modal interactive learning is an advanced data processing method that enhances the expressiveness of models by integrating information from different modalities (such as structured, semi-structured, and unstructured data). Furthermore, the proposed triple attention mechanism (modal, spatial, and temporal) deeply integrates multimodal data into a unified spatiotemporal graph structure while inheriting the physical constraints of the road network graph. This provides highly representative feature vectors for subsequent prediction and optimization, enhancing the model's generalization and robustness across both temporal and spatial dimensions.

[0110] In this embodiment, real-time prediction of public transportation conditions based on multimodal fusion feature vectors includes:

[0111] The spatiotemporal graph convolutional network (STGCN) is used to jointly extract spatiotemporal features from the fused features;

[0112] A multi-scale time window mechanism is used to capture temporal dependencies of different granularities;

[0113] The feature weights are dynamically adjusted through the spatiotemporal gating unit to output the public transportation flow prediction value.

[0114] Based on graph attention aggregation, this paper further uses multi-layer spatiotemporal graph convolution (ST-GCN) to extract deep spatiotemporal joint features, significantly improving the accuracy and timeliness of traffic flow prediction. At the same time, it combines the spatiotemporal gating unit to adaptively adjust the weights of each branch, improving the model's adaptability to complex traffic patterns.

[0115] In this embodiment, a public transportation scheduling strategy is formulated based on the particle swarm optimization algorithm, such as Figure 3 As shown, including:

[0116] Using the multimodal fusion feature vector as the initial position, randomly initialize the speed and position of each particle (representing different scheduling schemes, such as the number of vehicles at different time points, route selection, etc.);

[0117] Designing a comprehensive fitness function that combines factors such as vehicle cost, passenger waiting time, and load balancing can be:

[0118]

[0119] in, is the weight coefficient, which can be adjusted according to actual conditions; 、 、 They represent vehicle cost, passenger waiting time and route load balance respectively;

[0120] Calculate particle fitness based on public transportation flow prediction values;

[0121]

[0122] in, is the updated speed, is the velocity of particle n in dimension m, is the position of particle n in dimension m, is the shrinkage factor, and are the maximum and minimum values ​​of the inertia weight, t is the current number of iterations, is the maximum number of iterations, are the initial and final learning factors respectively The value of are the initial and final learning factors respectively The value of is the historical optimal position of particle n in the mth dimension, is the optimal position of the entire particle swarm in the mth dimension, and is a random number between [0, 1];

[0123]

[0124] in, is the updated location.

[0125] The update process takes into account spatiotemporal constraints, such as:

[0126] Maximum empty distance: limits the maximum distance a vehicle can travel without passengers.

[0127] Minimum departure interval: ensure that there is sufficient interval time between consecutive departures from the same station.

[0128] These constraints can be implemented by adding penalty terms to the fitness function. For example, if a particle violates the minimum departure interval constraint, a large penalty value can be added to its fitness.

[0129] The iterative update process is terminated when the maximum number of iterations is reached or the fitness change is less than the set threshold, and the global optimal solution is obtained. The global optimal solution is the public transportation scheduling strategy.

[0130] The inertia weight of the present invention controls the degree to which the original velocity of a particle is maintained during velocity updates. A larger inertia weight facilitates global search, while a smaller inertia weight facilitates local, refined search. The present invention adjusts the inertia weight in a linearly decreasing manner, effectively balancing the capabilities of global and local searches. Furthermore, to prevent excessive particle velocity from leading to search instability, the present invention introduces a contraction factor to limit velocity. This, combined with an improved particle swarm optimization (PSO) algorithm, aims to minimize operating costs, passenger waiting time, and load balancing, while satisfying a series of practical operational constraints (such as maximum idle distance, minimum departure interval, etc.) to develop an optimal bus scheduling solution. This effectively improves the performance of the PSO algorithm in public transportation vehicle scheduling problems.

[0131] Mapping the optimal solution to actual scheduling operations can include the following:

[0132] Departure time adjustment: Adjust the departure schedule of each line according to the departure time information in the optimal solution.

[0133] Vehicle quantity adjustment: Increase or decrease the number of vehicles put into operation based on the vehicle quantity information in the optimal solution.

[0134] Operating frequency adjustment: Adjust the departure intervals between vehicles based on the operating frequency information in the optimal solution.

[0135] Operation route adjustment: Adjust the vehicle's operation route based on the path information in the optimal solution, such as detouring around congested sections or adding temporary routes.

[0136] Station stop time adjustment: Optimize the vehicle's stay time at each station based on the station stop time information in the optimal solution.

[0137] This embodiment is not limited thereto, and the specific scheduling content can be dynamically adjusted according to actual needs.

[0138] This embodiment provides a computer device, including: a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, a public transportation scheduling optimization method driven by multimodal time series data analysis is implemented.

[0139] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, a public transportation scheduling optimization method driven by multimodal time series data analysis is implemented.

[0140] Those skilled in the art will appreciate that all or part of the steps of the above-mentioned method embodiments may be implemented by hardware associated with program instructions, and the aforementioned program may be stored in a computer-readable storage medium. When the program is executed, the program executes the steps of the above-mentioned method embodiments. The aforementioned storage medium includes various media that can store program codes, such as mobile storage devices, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0141] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0142] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A public transportation scheduling optimization method driven by multimodal time series data analysis, characterized in that: include: Acquire multimodal traffic data and perform spatiotemporal alignment preprocessing to extract feature vectors; Fusion of multiple feature vectors based on cross-modal interactive learning method; Real-time prediction of public transportation conditions is performed based on multimodal fusion feature vectors to obtain public transportation flow prediction values; Formulate public transportation dispatch strategies based on public transportation flow forecasts; Multimodal traffic data includes vehicle-mounted satellite positioning data, traffic flow data, and real-time data from social media apps. Structured data includes traffic flow data, semi-structured data includes vehicle-mounted satellite positioning data, and unstructured data includes text data from social media apps. Extracting feature vectors includes: Processing structured data based on autoencoders; Construct an urban traffic network map and extract spatial features from semi-structured data based on a graph attention network. The adjacency matrix of the graph attention network is constructed based on the road topology. Perform time series analysis and key information extraction on unstructured data based on bidirectional long short-term memory network; Based on the cross-modal interactive learning method, multiple feature vectors are integrated, including: (1) A multimodal spatiotemporal correlation graph is constructed based on the urban traffic network graph. The nodes represent the feature vectors of different modalities, and the edge weights are calculated through the spatiotemporal attention mechanism. The edge weight calculation process is as follows: Calculate modality attention weights: Among them, W Q , W K They represent the query transformation matrix and the key transformation matrix respectively, K represents the number of attention heads, D is the feature dimension after transformation, and h i and h j Represent the feature vectors of node i and node j respectively, α modal is the modality attention weight; Calculate the attention weight based on spatial distance: Among them, d geo represents the Euclidean distance of the grid center point, σ is the time attenuation coefficient, α space Represents the attention weight calculated based on spatial distance; Calculate the attention weight based on the time difference: Among them, τ represents the time decay coefficient, Δt represents the time interval between two nodes, α time Represents the attention weight calculated based on the time difference; Normalized attention weights: w i→j =softmax(a modal +a space +a time ); Among them, softmax represents the activation function, w i→j represents the final attention weight from node i to node j; (2) Aggregating multimodal features through graph attention networks, including: Use K independent attention heads to parallelize the computation: in, represents the value transformation matrix of the kth attention head, represents the neighbor nodes of node i, represents the normalized attention weight of the k-th attention head, Represents the features of the kth attention head; After parallel calculation of multiple attention heads, multi-head feature splicing is performed to output intermediate fusion features; Perform residual connection and normalization on the intermediate fusion features; (3) After the multimodal features are aggregated by the graph attention network, the node features are output through multi-layer spatiotemporal graph convolution. The node features are mapped into multimodal fusion feature vectors through the fully connected layer.

2. The public transportation scheduling optimization method driven by multimodal time series data analysis according to claim 1 is characterized in that: Real-time prediction of public transportation conditions based on multimodal fusion feature vectors, including: Use the spatiotemporal graph convolutional network to jointly extract spatiotemporal features from the fusion features; A multi-scale time window mechanism is used to capture temporal dependencies of different granularities; The feature weights are dynamically adjusted through the spatiotemporal gating unit to output the public transportation flow prediction value.

3. The public transportation scheduling optimization method driven by multimodal time series data analysis according to claim 1 is characterized in that: Formulate public transportation scheduling strategies based on particle swarm optimization algorithm, including: Use the multimodal fusion feature vector as the initial position and define the adaptability function; Calculate particle fitness based on public transportation flow prediction values; Update the position and velocity of each particle according to its fitness, taking into account the spatiotemporal constraints during the update process; The update process is terminated when it reaches the maximum number of iterations or the fitness change is less than the set threshold, and the global optimal solution is obtained. The global optimal solution is the public transportation scheduling strategy.

4. The public transportation scheduling optimization method driven by multimodal time series data analysis according to claim 3 is characterized in that: The update formula for the position and velocity of each particle is: Among them, v n,m (t+1) is the updated velocity, v n,m (t) is the velocity of particle n in the m dimension, x n,m (t) is the position of particle n in the m dimension, X is the shrinkage factor, ω max and ω min are the maximum and minimum values ​​of the inertia weight, t is the current iteration number, T max is the maximum number of iterations, c 1n ,c 1f are the initial and final values ​​of the learning factor c1, c 2n ,c 2f are the initial and final values ​​of the learning factor c2, pbest n,m is the historical optimal position of particle n in the mth dimension, gbest m is the optimal position of the entire particle swarm in the mth dimension, r1(t) and r2(t) are random numbers between [0, 1]; x n,m (t+1)=x n,m (t)+v n,m (t+1); Among them, x n,m (t+1) is the updated position.

5. The public transportation scheduling optimization method driven by multimodal time series data analysis according to claim 1 is characterized in that: Public transport scheduling strategies include: Vehicle departure time: the time when each vehicle departs from the departure station; Number of vehicles: the number of vehicles required on a certain route; Frequency of vehicle operation: the interval between vehicles; Vehicle operation path: whether the vehicle needs to detour certain sections of road or adjust the operation route; Stop time at each station: the time the vehicle stays at each station.

6. A computer device, characterized in that: include: A memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the method according to any one of claims 1 to 5 is implemented.

7. A computer-readable storage medium, characterized in that The storage medium stores a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 5.

Images