Method and system for optimizing urban vehicle carrying capacity based on hybrid prediction

Abstract

The application provides a kind of urban vehicle carrying capacity scale optimization method and system based on mixed prediction, it is related to traffic transport engineering technical field, the method steps include obtaining city, traffic, running track, travel and point of interest data, and pre-processing;Supply capacity and travel demand mixed two-way prediction model is constructed, based on the upper limit of physical capacity of urban road and parking resources, combined with the reasonable sharing rate of taxicab in traffic structure to calculate the supply capacity, by fusing deep learning and traffic planning theory to simulate the space-time network distribution of origin-destination travel, obtain the initial carrying capacity scale based on travel demand;The initial carrying capacity scale is corrected based on the model of multimodal traffic development trend and estimated and adjusted based on horizontal city comparison, to obtain the recommended carrying capacity scale after optimization;Set up multi-dimensional monitoring index system, real-time monitoring and evaluation of the operation state of taxicab market, execute corresponding control response measures.

Description

Technical Field

[0001] This invention relates to the field of transportation engineering technology, and in particular to a method and system for optimizing urban vehicle capacity based on hybrid prediction. Background Technology

[0002] The rational allocation of urban vehicle capacity is key to ensuring efficient travel between origin and destination, optimizing transportation resource allocation, and promoting the sustainable development of the industry. With accelerating urbanization and increasingly diversified transportation demands, taxis are playing an increasingly important role in urban public transportation systems.

[0003] However, determining the scale of transportation capacity involves multiple complex factors such as the city's development level, population distribution, travel preferences, road resource carrying capacity, vehicle operating efficiency, and market supply and demand. Traditional capacity planning methods often rely on single-dimensional experience judgments or simplified statistical models, which are usually limited by the channels and accuracy of data acquisition. This results in a large deviation between the predicted development trend of ride-hailing scale and the actual situation, and a mismatch between capacity allocation and transportation demand. It is difficult to fully capture the dynamics and complexity of the urban transportation system, which often leads to a mismatch between capacity allocation and residents' actual travel needs. This can either result in excess capacity and wasted resources, or cause supply and demand imbalance and travel difficulties. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for optimizing urban vehicle capacity based on hybrid prediction, in order to solve the problems mentioned in the background art, where traditional capacity planning methods often rely on single-dimensional empirical judgments or simplified statistical models, making it difficult to fully capture the dynamics and complexity of urban transportation systems, and easily leading to problems such as excess capacity, waste of resources, supply and demand imbalance, and travel difficulties.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for optimizing urban vehicle capacity based on hybrid prediction, comprising the following steps: acquiring urban basic data, traffic operation data, taxi operation trajectory data, origin-destination travel data, and point of interest (POI) data, and preprocessing them to obtain a multidimensional spatiotemporal dataset; based on the multidimensional spatiotemporal dataset, constructing a hybrid bidirectional prediction model of supply capacity and travel demand, calculating supply capacity based on the physical capacity limit of urban roads and parking resources, combined with the reasonable modal share of taxis in the traffic structure, simulating the spatiotemporal network distribution of origin-destination travel by integrating deep learning and traffic planning theory, and obtaining an initial capacity scale based on travel demand; sequentially performing model correction based on multimodal traffic development trends and estimation adjustment based on horizontal city comparisons on the initial capacity scale to obtain an optimized recommended capacity scale; setting a multidimensional monitoring indicator system based on the optimized recommended capacity scale to monitor and evaluate the taxi market operation status in real time, and triggering graded early warnings and implementing corresponding regulatory response measures according to the evaluation results, so as to achieve a dynamic balance between supply and demand of capacity.

[0006] Optionally, the preprocessing steps specifically include: cleaning the collected vehicle trajectory data, filtering valid data according to the set urban area latitude and longitude range, removing null values, abnormal passenger status attribute values, and abnormal trajectory offset data, and performing dimensionality reduction on the data; extracting operational feature parameters from the cleaned trajectory data, including ride-hailing demand, demand time distribution, passenger carrying time distribution, passenger search time distribution, and mileage utilization rate; reclassifying the travel interest point (POI) data, merging similar categories and removing categories with low correlation to travel demand, and performing spatial grid filtering based on the set urban area to calculate the interest point density and interest point diversity index of each grid unit.

[0007] Optionally, the step of calculating the supply capacity based on the physical capacity limit of urban roads and parking resources, combined with the reasonable modal share of taxis in the traffic structure, specifically includes: calculating the taxi capacity that road resources can support by using the mileage of urban roads at all levels, the capacity per unit length, the average operating speed of motor vehicles, the vehicle utilization rate during peak hours, and the modal share of taxis in all modes of travel; assessing the upper limit of the number of taxis that parking resources can support by using the total supply of urban parking spaces, the proportion of taxi parking demand, and the turnover efficiency of parking spaces; deducing the capacity scale that matches the city's operational efficiency by setting a reasonable empty-running rate range for taxis, combined with the average daily operating time and average operating speed of vehicles; and determining the taxi capacity scale based on the calculation results.

[0008] Optionally, the step of simulating the spatiotemporal network distribution of origin-destination travel by integrating deep learning and traffic planning theory specifically includes: constructing a hybrid prediction model that integrates a long short-term memory network and a four-stage traffic planning method; in the travel generation stage of the hybrid prediction model, using a long short-term memory network to process the temporal characteristics of historical travel volume, order volume, order temporal change rate, and points of interest data to predict the total travel demand of each grid area; in the travel distribution stage, using a convolutional long short-term memory network model to divide the urban space into uniform grids, and using historical order demand, demand change rate, points of interest density, and road network connectivity as multi-channel spatiotemporal grid data as input, and outputting the predicted travel demand of each grid area in the future period.

[0009] Optionally, the model correction step based on multimodal transportation development trends specifically includes: selecting multimodal transportation factors affecting the scale of taxi capacity, wherein the multimodal transportation factors include at least: the market penetration rate of driverless taxis, the deployment scale of shared electric bicycles, the penetration rate of low-altitude passenger transport, the growth rate of private motor vehicle ownership, and changes in the public transport modal share; assigning weight coefficients to each of the multimodal transportation factors, and, based on the rate of change of each influencing factor in future time periods, combined with the corresponding weight coefficients, weighting and correcting the initial capacity scale obtained based on travel demand to obtain the capacity scale corrected by the multimodal transportation factors.

[0010] Optionally, the steps of estimation and adjustment based on horizontal city comparison specifically include: selecting multiple cities similar to the target city in terms of scale, economic development stage, or urban function as comparison cities; extracting a set of common indicators affecting taxi ownership from the target city and each comparison city; processing the common indicators using principal component analysis to determine the influence weight of each indicator; and performing weighted calculations based on the ratios of the target city and each comparison city on each indicator, the actual taxi ownership of each comparison city, and a preset similarity weight, to obtain a reference value for transport capacity based on horizontal comparison.

[0011] Optionally, the steps of real-time monitoring and evaluation of the taxi market operation status, and triggering tiered early warnings and implementing corresponding regulatory response measures based on the evaluation results, specifically include: continuous monitoring and evaluation based on a preset indicator system covering four dimensions: supply and demand balance, compliance governance, operational efficiency, and safety risks; setting three-level quantitative early warning thresholds (yellow, orange, and red) for the core indicators in the indicator system; when the monitoring data triggers a yellow warning, implementing response measures including generating a market operation monitoring report, issuing warnings and guidance to operating platforms with low compliance rates, and releasing compliant ride guidance information to the public; when the monitoring data triggers an orange warning, implementing response measures including implementing economic regulation on non-compliant orders, providing subsidies to compliant drivers providing services during peak hours, and optimizing the platform's order scheduling strategy through algorithms; and when the monitoring data triggers a red warning, implementing response measures including suspending new capacity access permits, restricting or shutting down the service functions of low-compliance platforms in specific areas, and organizing joint enforcement actions across administrative departments.

[0012] On the other hand, the present invention also provides an urban vehicle capacity optimization system based on hybrid prediction, comprising: an acquisition module for acquiring urban basic data, traffic operation data, taxi operation trajectory data, origin-destination travel data, and point of interest (POI) data, and preprocessing them to obtain a multidimensional spatiotemporal dataset; a model building module for constructing a hybrid bidirectional prediction model of supply capacity and travel demand based on the multidimensional spatiotemporal dataset, calculating supply capacity based on the physical capacity limit of urban roads and parking resources, combined with the reasonable share of taxis in the traffic structure, simulating the spatiotemporal network distribution of origin-destination travel by integrating deep learning and traffic planning theory, and obtaining an initial capacity scale based on travel demand; an optimization module for sequentially performing model correction based on multimodal traffic development trends and estimation adjustment based on horizontal city comparisons on the initial capacity scale to obtain an optimized recommended capacity scale; and an early warning and control module for setting a multidimensional monitoring indicator system based on the optimized recommended capacity scale, monitoring and evaluating the taxi market operation status in real time, and triggering graded early warnings and implementing corresponding control response measures according to the evaluation results to achieve a dynamic balance between supply and demand of capacity.

[0013] On the other hand, the present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method for optimizing urban vehicle capacity based on hybrid prediction.

[0014] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method for optimizing urban vehicle capacity based on hybrid prediction.

[0015] Compared with the prior art, the beneficial effects of the present invention are: This application overcomes the limitations of traditional single-dimensional forecasting methods by constructing a hybrid two-way forecasting model that integrates supply capacity and travel demand, significantly improving the scientific rigor and accuracy of capacity calculation. Furthermore, through a dual optimization mechanism of multimodal transportation development trend correction and horizontal city comparison estimation adjustment, the foresight and realistic rationality of the planning results are enhanced. Finally, relying on a management closed loop formed by a multi-dimensional monitoring indicator system and tiered early warning and control response, a fundamental shift from static planning to dynamic balance in capacity allocation is achieved, thus systematically solving the industry problem of capacity-demand mismatch and improving the overall utilization efficiency of urban transportation resources and the level of refinement in industry governance. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the method steps of the present invention.

[0017] Figure 2 This is a flowchart of the method of the present invention.

[0018] Figure 3 This is a schematic diagram of the system structure of the present invention.

[0019] In the diagram: 10 - Acquisition module, 20 - Model building module, 30 - Optimization module, 40 - Early warning and control module. Detailed Implementation

[0020] The present invention will now be clearly and completely described in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0021] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be used interchangeably where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0022] Those skilled in the art will understand that, unless explicitly stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in the specification of this application means the presence of features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0023] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0024] It should be understood that the sequence number and size of each step in this embodiment do not imply the order of execution. The execution order of each process is determined by its function and internal logic, and should not constitute any limitation on the implementation process of this application embodiment.

[0025] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0026] Please refer to Figures 1-2This invention discloses a method for optimizing urban vehicle transport capacity based on hybrid prediction. The steps include: acquiring urban basic data, traffic operation data, taxi operation trajectory data, origin-destination travel data, and points of interest (POI) data, and preprocessing them to obtain a multidimensional spatiotemporal dataset; constructing a hybrid bidirectional prediction model of supply capacity and travel demand based on the multidimensional spatiotemporal dataset; calculating supply capacity based on the physical capacity limits of urban roads and parking resources, combined with the reasonable modal share of taxis in the traffic structure; simulating the spatiotemporal network distribution of origin-destination travel by integrating deep learning and traffic planning theories to obtain an initial transport capacity based on travel demand; sequentially revising the initial transport capacity based on multimodal traffic development trends and adjusting the estimation based on horizontal city comparisons to obtain an optimized recommended transport capacity; setting a multidimensional monitoring indicator system based on the optimized recommended transport capacity to monitor and evaluate the taxi market operation status in real time, and triggering tiered early warnings and implementing corresponding control response measures based on the evaluation results to achieve a dynamic balance between transport capacity supply and demand.

[0027] Specifically, the system collects basic urban data such as GDP, population, and road mileage from multiple sources, including the Municipal Transportation Bureau, statistical yearbooks, Gaode Maps open platform, and ride-hailing regulatory platforms. It also gathers traffic operation data such as vehicle GPS trajectories and order records, origin-destination travel survey data, and Points of Interest (POI) data including names, categories, and latitude / longitude. This data is then uniformly cleaned and formatted to construct a cross-domain, multi-dimensional spatiotemporal dataset. Next, a hybrid two-way forecasting process is performed based on this dataset: on the supply side, the capacity is calculated from the perspective of the upper limit of urban physical infrastructure supply, based on road capacity and reasonable empty-running rate constraints. On the demand side, a model integrating deep learning and transportation planning theory is constructed to simulate future transportation demand, obtaining an initial capacity value based on demand. Then, this initial value undergoes a two-stage optimization: first, considering the impact of future transportation modes such as autonomous driving and shared electric bicycles, the model is revised based on multimodal transportation development trends; second, similar cities are selected for horizontal comparison, and a reference value is obtained through principal component analysis and weighted calculation. Finally, the optimized recommended capacity scale is derived. Finally, the system continues to operate, and based on the recommended capacity scale, it sets indicators covering four dimensions: supply and demand, compliance, efficiency, and safety for real-time monitoring. Once an indicator exceeds the set yellow, orange, and red thresholds, it automatically triggers corresponding control measures from information guidance to administrative intervention, thus forming a dynamic management closed loop of prediction-optimization-monitoring-control to achieve a continuous balance between capacity supply and demand.

[0028] This application overcomes the limitations of traditional single-dimensional forecasting methods by constructing a hybrid two-way forecasting model that integrates supply capacity and travel demand, significantly improving the scientific rigor and accuracy of capacity calculation. Furthermore, through a dual optimization mechanism of multimodal transportation development trend correction and horizontal city comparison estimation adjustment, the foresight and realistic rationality of the planning results are enhanced. Finally, relying on a management closed loop formed by a multi-dimensional monitoring indicator system and tiered early warning and control response, a fundamental shift from static planning to dynamic balance in capacity allocation is achieved, thus systematically solving the industry problem of capacity-demand mismatch and improving the overall utilization efficiency of urban transportation resources and the level of refinement in industry governance.

[0029] In some embodiments, the preprocessing steps specifically include: cleaning the collected vehicle trajectory data, filtering valid data according to the set urban area latitude and longitude range, removing null values, abnormal passenger status attribute values, and abnormal trajectory offset data, and performing dimensionality reduction on the data; extracting operational feature parameters from the cleaned trajectory data, the feature parameters including ride-hailing demand, demand time distribution, passenger carrying time distribution, passenger search time distribution, and mileage utilization rate; reclassifying the travel point of interest (POI) data, merging similar categories and removing categories with low correlation to travel demand, and performing spatial grid filtering based on the set urban area, calculating the POI density and POI diversity index of each grid unit.

[0030] Specifically, for vehicle trajectory data, spatial filtering is first performed based on city geographical boundaries to remove records outside the defined range. Data cleaning then follows, including deleting records with null values ​​in key fields, correcting or removing abnormal passenger status values, and identifying significant trajectory drift points caused by signal loss. To improve processing efficiency, Python tools are used to examine the data columns, removing redundant fields irrelevant to the analysis, thus achieving data dimensionality reduction.

[0031] From the cleaned data, the demand for ride-hailing services in each region and time period was calculated as follows: ;in, Let i be the number of orders in region i during time period t; For decision variables, the value is 1 if order k occurs in region i and time period t, and 0 otherwise; The average passenger load factor is set to 2, based on historical data and experience, since it is difficult to obtain the number of passengers per vehicle.

[0032] The time period distribution of ride-hailing demand is as follows: ;in, The percentage of orders in time period t; The total number of orders in time period t; This represents the total number of time periods.

[0033] Passenger travel time distribution is as follows: ; ;in, This is the average number of passengers. For passenger variance, The time spent on the kth order.

[0034] The time distribution for finding customers is as follows: The statistical method is similar to that for passenger search time. If the passenger search time exceeds 1 hour, it is classified as an outlier and needs to be removed.

[0035] For POI data, the original 23 categories are merged and filtered according to their functional attributes. For example, categories that are strongly related to travel, such as shopping and healthcare, are retained, while categories with low correlation, such as gas stations, are removed. Spatial filtering is also performed within the same study area.

[0036] Finally, the POI density is calculated for each geographic grid cell. and POI diversity index Linking Points of Interest (POIs) with Travel Demand ;in, For POI density, For the POI diversity index, The total number of point of interest types. The percentage of the number of interest points of type j. Let i be the number of POIs of type i. The area of ​​the region where the POI is located. Let be the travel demand of the i-th grid cell. For the regression constant term, The number of interest points of type j within the i-th grid cell. For the POI attractiveness weight of type i, This is a constant term.

[0037] This application improves data quality and processing efficiency by specifically cleaning, filtering, and feature engineering multi-source heterogeneous data, and by cleaning and reducing the dimensionality of trajectory data. The extracted operational feature parameters directly quantify the market operation status, providing key inputs for the model. The reclassification and gridded computation of POI data enables a quantitative representation of the urban functional spatial structure. By transforming the raw data into a standardized, gridded, and computable multidimensional spatiotemporal dataset, the problem of messy and low information density of the raw data is effectively overcome, ensuring the accuracy and efficiency of subsequent analysis.

[0038] In some embodiments, the step of calculating the supply capacity based on the physical capacity limit of urban roads and parking resources, combined with the reasonable modal share of taxis in the traffic structure, specifically includes: calculating the taxi capacity scale that road resources can support by using the mileage of urban roads at all levels, the capacity per unit length, the average operating speed of motor vehicles, the vehicle utilization rate during peak hours, and the modal share of taxis in all modes of travel; assessing the upper limit of the number of taxis that parking resources can support by using the total supply of urban parking spaces, the proportion of taxi parking demand, and the turnover efficiency of parking spaces; inversely calculating the capacity scale that matches the city's operational efficiency by setting a reasonable empty-running rate range for taxis, combined with the average daily operating time and average operating speed of vehicles; and determining the taxi capacity scale based on the calculation results.

[0039] Specifically, the supply capacity assessment includes three parallel evaluation paths. The first is a road resource-based assessment: obtaining the total mileage of urban expressways, arterial roads, and secondary arterial roads; referencing standards to determine the capacity per unit length as 800 vehicles / hour; using floating car data to determine the average speed of motor vehicles as approximately 24 km / h; obtaining data from regulatory platforms that the peak usage rate of ride-hailing services is approximately 44%; and setting the ride-hailing travel proportion at 5%-6.8% based on the principle of moderate development; and substituting these values ​​into the formula to calculate the scale. Where Q represents the reasonable total scale; L represents the mileage of roads at all levels in the city; C represents the traffic capacity per unit length; V represents the average operating speed of motor vehicles in the city; γ represents the current peak usage rate of ride-hailing services in the city, that is, the proportion of ride-hailing services operating during peak hours to the total number of ride-hailing services in the city; and α represents the proportion of ride-hailing trips among all modes of transportation in the city.

[0040] Second, based on the calculation of parking resources: survey the total number of public parking spaces in the city, estimate the proportion of parking demand and parking space turnover rate of taxis in combination with the characteristics of taxi operation, and assess the upper limit of the scale that can be supported.

[0041] Third, an assessment based on a reasonable empty-running rate: Referring to industry experience, the reasonable empty-running rate K is set in the range of 30%-40%, combined with the average daily operating time T, average operating speed V, and number of vehicles n, using the formula... To determine the optimal scale of taxi fleet while maintaining a certain empty-running rate, we can deduce the following: K represents the empty-running rate; T represents the average daily operating time of a typical taxi; V represents the average operating speed of a taxi; and n represents the number of taxis. Based on the experience of advanced taxi operations in some developed cities in China and relevant domestic and international analyses and surveys, combined with GPS data analysis of taxis in Jinan, a city-wide empty-running rate of 30% to 40% is considered reasonable. An excessively high empty-running rate leads to resource waste, while an excessively low rate results in difficulty hailing a taxi, reducing passenger satisfaction. Therefore, maintaining an empty-running rate of around 35% for a typical taxi fleet will achieve the best results.

[0042] Ultimately, when making decisions, the final reference value for the supply capacity dimension is determined based on the calculation results of the three factors to ensure that it does not exceed the rigid constraints of urban resources.

[0043] This application comprehensively assesses three complementary physical or efficiency constraints: road resources, parking resources, and reasonable empty-running rate. This approach overcomes the limitations of relying solely on single resource limits. Road resource assessment reflects dynamic traffic capacity constraints, parking resource assessment reflects static capacity constraints, and the reasonable empty-running rate assessment incorporates industry-relevant values ​​for operational economics and service efficiency. This multi-dimensional approach to determining the final supply capacity reference value ensures that the planned capacity does not exceed the limits of urban infrastructure and reasonable operational efficiency, thus enhancing the feasibility and safety of the planning scheme.

[0044] In some embodiments, the step of simulating the spatiotemporal network distribution of origin-destination travel by integrating deep learning and traffic planning theory specifically includes: constructing a hybrid prediction model that integrates a long short-term memory network and a four-stage traffic planning method; in the travel generation stage of the hybrid prediction model, using a long short-term memory network to process the temporal characteristics of historical travel volume, order volume, order temporal change rate, and point of interest data to predict the total travel demand of each grid area; in the travel distribution stage, using a convolutional long short-term memory network model to divide the urban space into uniform grids, and using historical order demand, demand change rate, point of interest density, and road network connectivity as multi-channel spatiotemporal grid data as input, and outputting the predicted travel demand of each grid area in the future time period.

[0045] Specifically, traffic distribution forecasting refers to converting the total occurrence and attraction volumes of traffic zones in traffic activity forecasting into specific travel volumes for each traffic zone. In this process, different methods are used to simulate and predict the travel distribution between traffic zones. By considering the cross-traffic between intersections or nodes, the traffic distribution on different paths is estimated. The calculation formula is: ; ; In the formula: This indicates the traffic volume between area i and area j in the coming years; This represents the traffic volume between region i and region j in the base year; This indicates the multiple by which traffic volume increases in zone i; This indicates the multiple by which traffic volume increases in zone j.

[0046] In the traffic distribution prediction stage, ConvLSTM is used for prediction. This model combines a convolutional neural network (CNN) and a long short-term memory network (LSTM) deep learning model, which can be used to process sequential data with spatiotemporal dependencies, such as vehicle trajectories. Forget gate: Input gate: Candidate state: Cell state: Output gate: ; Latent state: Where ∗ represents convolution operation and ⊙ represents element-wise multiplication; Ht-1 is the input of the current time step; Ht-1 is the hidden state of the previous time step. These are the convolution kernel weights. This is the bias term. σ uses the Sigmoid activation function to gate the signal.

[0047] The city was divided into uniform grids, and data was aggregated in fixed 30-minute time windows. GPS trajectories of taxi orders within the past 12 hours were converted into spatiotemporal grid data, and a gridded demand heatmap was created. The urban development baseline data from previous surveys were incorporated as input.

[0048] The order quantity in the current time slice is selected as the primary demand channel, while the demand change rate, POI density, and road network connectivity are used as other demand channels. For channel set, The main channel of demand is The demand change rate channel is POI density channel is The road network connectivity channel is n is the number of intersections, and L is the road length.

[0049] To better reconstruct travel characteristics, spatiotemporal context features are further integrated on top of basic features. This paper adopts a staged fusion approach: in ConvLSTMCELL1, 3×3 basic convolutions are used to extract low-order grid features; in the middle layer of ConvLSTMCELL2, 5×5 convolutions are used in the time dimension to extract hourly features. ,Week and special holidays Predict separately and dynamically weight using a gating mechanism; calculate the neighborhood mean or maximum value using a 3×3 convolutional kernel in the spatial dimension; temporal context-temporal feature embedding: Spatial context-neighborhood statistics: In ConvLSTMCELL3, a 5×5 convolution is used to introduce a global spatiotemporal context by predicting urban traffic flow.

[0050] Parallel use of 3×3 and 5×5 convolutional kernels of different sizes captures multi-scale spatial patterns. In ConvLSTMCELL2, parallel multi-scale convolution increases the number of channels, with 64-channel concatenations increasing the number to 128. The lower layer captures the short-term dependence of ride-hailing trajectory data, while the higher layer captures the long-term trend of ride-hailing predictions in urban traffic flow and economic growth.

[0051] Attention mechanisms are divided into spatial attention and temporal attention. Spatial attention... Enhance attention by focusing on POI hotspot areas Time-based attention enhancement is achieved through key morning and evening rush hours and special holidays. . ; .

[0052] In traffic demand forecasting, the urban traffic volume is obtained, with ride-hailing services accounting for 5%–6.8% of all transportation trips. The generation and attraction of ride-hailing services between residential areas are calculated. Where: C represents the reasonable scale of ride-hailing services; For ride-hailing transportation demand, p represents the daily order volume, and p represents the total vehicle dispatch rate. This is the surplus factor, which is taken as 0 under the baseline condition.

[0053] This application addresses the bottleneck of traditional methods in predicting spatiotemporally distributed travel demand by embedding Long Short-Term Memory (LSTM) networks, which excel at capturing complex temporal dependencies, into the framework of the classic four-stage transportation planning method. In the travel generation stage, LSTM is used to process historical sequences and related factors, improving the prediction accuracy of the nonlinear changes in total demand over time. In the travel distribution stage, a ConvLSTM model is used to process multi-channel spatiotemporal grid data, enabling simultaneous learning of the spatial clustering effect and temporal evolution of travel demand. This achieves accurate, end-to-end prediction of fine-grained grids and future travel distribution, significantly improving the spatial specificity and temporal accuracy of capacity calculation based on travel demand.

[0054] In some embodiments, the model correction step based on multimodal transportation development trends specifically includes: selecting multimodal transportation factors affecting the scale of taxi capacity, wherein the multimodal transportation factors include at least: the market penetration rate of driverless taxis, the deployment scale of shared electric bicycles, the penetration rate of low-altitude passenger transport, the growth rate of private motor vehicle ownership, and changes in the public transport modal share; assigning weight coefficients to each of the multimodal transportation factors, and, based on the rate of change of each influencing factor in future time periods, combined with the corresponding weight coefficients, weighting and correcting the initial capacity scale obtained based on travel demand to obtain the capacity scale corrected by the multimodal transportation factors.

[0055] Specifically, this method provides a calibration anchor point based on real-world experience for the local model prediction results. The corrected model uses a multiple linear regression model, and the weights are determined by expert methods. The correction formula is as follows: In the formula: , This represents the current and revised forecast values. The weights of each factor. The rate of change for each factor.

[0056] This application incorporates key variables of future transportation structure transformation into current capacity planning considerations, achieving a leap from reflecting the current situation to adapting to the future. Through pre-set impact weighting coefficients, it quantitatively assesses the potential diversion effect of emerging transportation modes such as autonomous driving, shared electric bicycles, and low-altitude passenger transport on taxi demand, as well as their sensitivity to trends such as motor vehicle growth and public transportation upgrades. This modified model essentially adds a forward-looking attenuation factor to capacity forecasting, making the planning results more strategically resilient and adaptable. It avoids the risk of plans becoming rapidly outdated or investments being wasted due to technological or model changes, providing a forward-looking basis for scientific decision-making.

[0057] In some embodiments, the step of estimation and adjustment based on horizontal city comparison specifically includes: selecting multiple cities similar to the target city in terms of scale, economic development stage, or urban function as comparison cities; extracting a set of common indicators affecting taxi ownership from the target city and each comparison city; processing the common indicators using principal component analysis to determine the influence weight of each indicator; and performing weighted calculations based on the ratios of the target city and each comparison city on each indicator, the actual taxi ownership of each comparison city, and a preset similarity weight to obtain a reference value for transport capacity based on horizontal comparison.

[0058] Specifically, this model primarily employs a horizontal city comparison estimation method to compare data on the number of taxis per 10,000 people, the number of ride-hailing vehicles per 10,000 people, daily order volume, ride-hailing capacity saturation in cities, and data from small and medium-sized cities with the current study city, thereby deriving a revised size prediction value. The calculation formula is as follows: In the formula: The value represents the estimated taxi ownership in the target city; j represents the number of comparison factors; n represents the number of comparison cities. Factors influencing the number of taxis; To compare the values ​​of the j-th factor in city i; This represents the value of the j-th factor in the target city. To compare the number of taxis in the city; To compare the similarity weights between the city and the target city.

[0059] This application uses principal component analysis to extract core common factors influencing transport capacity from numerous city indicators, avoiding the bias of subjective indicator selection. By calculating similarity weights, it ensures that samples more comparable to the target city have a greater impact. By placing the target city in a comparable coordinate system and using real development data from similar cities to validate and adjust the model results, it effectively prevents distorted results from complex models due to local data or parameter biases, enhancing the realistic rationality and decision-acceptability of the final recommended transport capacity scale.

[0060] In some embodiments, the steps of real-time monitoring and evaluation of the taxi market operation status, and triggering tiered early warnings and implementing corresponding regulatory response measures based on the evaluation results, specifically include: continuously monitoring and evaluating an indicator system covering four dimensions—supply and demand balance, compliance governance, operational efficiency, and safety risks—based on the optimized recommended capacity scale; setting three-level quantitative early warning thresholds (yellow, orange, and red) for the core indicators in the indicator system; when monitoring data triggers a yellow warning, implementing response measures including generating a market operation monitoring report, issuing warnings and guidance to operating platforms with low compliance rates, and releasing compliant ride guidance information to the public; when monitoring data triggers an orange warning, implementing response measures including implementing economic regulation on non-compliant orders, providing subsidies to compliant drivers providing services during peak hours, and optimizing the platform's order scheduling strategy through algorithms; and when monitoring data triggers a red warning, implementing response measures including suspending new capacity access permits, restricting or shutting down the service functions of low-compliance platforms in specific areas, and organizing joint enforcement actions across administrative departments.

[0061] Specifically, the system's indicator system, based on the optimized recommended capacity scale, includes four dimensions: supply and demand balance (core indicator: capacity saturation index, i.e., daily average number of orders / number of compliant vehicles), compliance governance (core indicator: proportion of illegally operating vehicles), operational efficiency (core indicator: empty mileage ratio), and safety risk (core indicator: accident rate per 10,000 kilometers). Each core indicator has clearly defined yellow, orange, and red thresholds. For example, a capacity saturation index exceeding 20 triggers a yellow alert, exceeding 24 triggers an orange alert, and exceeding 25 triggers a red alert. The system connects to real-time data streams and continuously calculates these indicators. When an indicator triggers a yellow alert, the system automatically generates a weekly report, sends risk warnings to platforms with low compliance rates, and pushes compliant ride reminders to citizens through the government affairs app. When an orange alert is triggered, the system instructs the regulatory platform to impose adjustment fees on non-compliant orders, while simultaneously issuing subsidies to compliant driver accounts that accept orders during peak hours, and triggering the platform's dispatch algorithm to enter "efficiency-first" mode. When a red alert is triggered, the system automatically freezes the online application process for new capacity permits, sends a service suspension order to platforms in areas where illegal operations account for more than 80%, and generates an enforcement task list that is synchronized to the joint enforcement system of transportation and public security. Data on the effectiveness of all response measures will be re-integrated into the real-time database, initiating the next round of monitoring and evaluation.

[0062] This application, through a pre-set four-dimensional indicator system and three-level quantitative thresholds, achieves automatic, real-time diagnosis and early warning classification of market operation status (such as supply and demand imbalance, insufficient compliance, inefficiency, and increased risk). More importantly, it matches each early warning level with a tiered, combined response measure package, ranging from information guidance and economic incentives to administrative enforcement, ensuring that early warnings can be automatically and rapidly transformed into actionable and differentiated regulatory actions. This completely changes the traditional management model of reactive, ex-post discovery, enabling proactive, precise, and efficient intervention in the transportation capacity market and ensuring the sustainable achievement of dynamic balance goals.

[0063] Please refer to Figure 3On the other hand, the present invention also provides an urban vehicle capacity optimization system based on hybrid prediction, comprising: an acquisition module 10, used to acquire urban basic data, traffic operation data, taxi operation trajectory data, origin-destination travel data, and point of interest (POI) data, and preprocess them to obtain a multidimensional spatiotemporal dataset; and a model building module 20, used to construct a hybrid bidirectional prediction model of supply capacity and travel demand based on the multidimensional spatiotemporal dataset, calculate the supply capacity based on the physical capacity limit of urban roads and parking resources, combined with the reasonable modal share of taxis in the traffic structure, and through fusion depth The system uses learning and transportation planning theory to simulate the spatiotemporal network distribution of origin-destination travel, obtaining an initial capacity scale based on travel demand. Optimization module 30 is used to sequentially modify the initial capacity scale using a model based on multimodal transportation development trends and to adjust estimations based on horizontal city comparisons, resulting in an optimized recommended capacity scale. Early warning and control module 40 is used to set a multi-dimensional monitoring indicator system based on the optimized recommended capacity scale, to monitor and evaluate the taxi market operation status in real time, and to trigger tiered early warnings and implement corresponding control response measures based on the evaluation results, thereby achieving a dynamic balance between capacity supply and demand.

[0064] On the other hand, the present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method for optimizing urban vehicle capacity based on hybrid prediction.

[0065] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method for optimizing urban vehicle capacity based on hybrid prediction.

[0066] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0067] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, database, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0068] The above are merely embodiments of the present invention and do not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention's specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for optimizing urban vehicle capacity based on hybrid prediction, characterized by the following steps: include: Acquire urban basic data, traffic operation data, transportation vehicle operation trajectory data, travel demand data, and travel interest point (POI) data, and preprocess them to obtain a multidimensional spatiotemporal dataset; Based on the multidimensional spatiotemporal dataset, a hybrid bidirectional prediction model of supply capacity and travel demand is constructed. Based on the physical capacity limit of urban roads and parking resources, and combined with the reasonable modal share of taxis in the traffic structure, the supply capacity is calculated. By integrating deep learning and traffic planning theory, the spatiotemporal network distribution of origin and destination travel is simulated to obtain the initial capacity scale based on travel demand. For the initial capacity scale, model correction based on multimodal transportation development trends and estimation adjustment based on horizontal city comparison are performed sequentially to obtain the optimized recommended capacity scale; Based on the optimized recommended capacity scale, a multi-dimensional monitoring indicator system is set up to monitor and evaluate the operation status of the taxi market in real time. According to the evaluation results, a graded early warning is triggered and corresponding regulatory response measures are implemented to achieve a dynamic balance between supply and demand of transportation capacity.

2. The method for optimizing urban vehicle capacity based on hybrid prediction according to claim 1, characterized in that, The preprocessing steps specifically include: The collected vehicle trajectory data is cleaned by filtering valid data according to the set urban area latitude and longitude range, removing null values, abnormal passenger status attribute values ​​and abnormal trajectory deviation data, and performing dimensionality reduction processing on the data. Operational characteristic parameters are extracted from the cleaned trajectory data. These characteristic parameters include ride-hailing demand, demand time distribution, passenger carrying time distribution, passenger search time distribution, and mileage utilization rate. The travel point of interest (POI) data is reclassified, similar categories are merged and categories with low relevance to travel demand are removed, and spatial grids are filtered based on the set urban areas to calculate the POI density and POI diversity index of each grid unit.

3. The method for optimizing urban vehicle capacity based on hybrid prediction according to claim 1, characterized in that, The steps for calculating supply capacity based on the physical capacity limit of urban roads and parking resources, combined with the reasonable modal share of taxis in the transportation structure, specifically include: The scale of taxi capacity that road resources can support is calculated by considering the mileage of roads at all levels in the city, the capacity per unit length, the average operating speed of motor vehicles, the vehicle utilization rate during peak hours, and the share of taxis in all modes of travel. The maximum number of taxis that parking resources can support is assessed by considering the total supply of urban parking spaces, the proportion of taxi parking demand, and the turnover efficiency of parking spaces. By setting a reasonable empty-running rate range for taxis, and combining the average daily operating time and average operating speed of vehicles, the capacity scale that matches the city's operational efficiency can be calculated. The scale of taxi capacity based on supply capacity is determined based on the calculation results.

4. The method for optimizing urban vehicle capacity based on hybrid prediction according to claim 1, characterized in that, The steps for simulating the spatiotemporal network distribution of origin-destination travel by integrating deep learning and traffic planning theory specifically include: Construct a hybrid prediction model that integrates long short-term memory networks and the four-stage traffic planning method; In the trip generation stage of the hybrid prediction model, a long short-term memory network is used to process the temporal characteristics of historical trip volume, order volume, order time-series change rate, and point of interest data to predict the total trip demand of each grid area. In the travel distribution phase, a convolutional long short-term memory network model is used to divide the urban space into uniform grids. Historical order demand, demand change rate, point of interest density, and road network connectivity are used as inputs to construct multi-channel spatiotemporal grid data, and the output is the travel demand prediction results for each grid area in the future period.

5. The method for optimizing urban vehicle capacity based on hybrid prediction according to claim 4, characterized in that, The steps for model correction based on multimodal traffic development trends specifically include: The multimodal transportation factors that influence the scale of taxi capacity are selected. These multimodal transportation factors include at least: the market penetration rate of driverless taxis, the deployment scale of shared electric bicycles, the penetration rate of low-altitude passenger transport, the growth rate of private motor vehicle ownership, and changes in the public transport modal share. Each of the multimodal traffic factors is assigned a weighting coefficient. Based on the rate of change of each influencing factor in the future time period, and combined with the corresponding weighting coefficient, the initial capacity scale obtained based on travel demand is weighted and corrected to obtain the capacity scale corrected by the multimodal traffic factors.

6. The method for optimizing urban vehicle capacity based on hybrid prediction according to claim 5, characterized in that, The steps for estimation and adjustment based on horizontal city comparison specifically include: Several cities that are similar to the target city in terms of size, stage of economic development, or urban function were selected as comparison cities. We extracted a set of common indicators affecting the number of taxis from the target city and the cities we compared with it. Principal component analysis was used to process the common indicators to determine the influence weight of each indicator. Based on the ratios of the target city and each comparative city in various indicators, the actual number of taxis in each comparative city, and combined with preset similarity weights, a weighted calculation is performed to obtain a reference value for the scale of transportation capacity based on horizontal comparison.

7. The method for optimizing urban vehicle capacity based on hybrid prediction according to claim 1, characterized in that, The steps for real-time monitoring and evaluation of the taxi market operation status, and for triggering tiered early warnings and implementing corresponding regulatory response measures based on the evaluation results, specifically include: Continuous monitoring and evaluation are conducted based on a pre-defined indicator system covering four dimensions: supply and demand balance, compliance governance, operational efficiency, and security risks. Three-level quantitative early warning thresholds (yellow, orange, and red) are set for the core indicators in the aforementioned indicator system; When monitoring data triggers a yellow alert, response measures will be implemented, including generating a market operation monitoring report, issuing warnings and guidance to operating platforms with low compliance rates, and releasing information on compliant ride-hailing to the public. When monitoring data triggers an orange alert, response measures will be implemented, including economic regulation of non-compliant orders, subsidies for compliant drivers providing services during peak hours, and optimization of the platform's order scheduling strategy through algorithms. When monitoring data triggers a red alert, response measures will be implemented, including suspending new capacity access permits, restricting or shutting down the service functions of low-compliance platforms in specific areas, and organizing joint enforcement actions across administrative departments.

8. A system for optimizing urban vehicle capacity based on hybrid prediction, characterized in that, include: The acquisition module is used to acquire urban basic data, traffic operation data, taxi operation trajectory data, origin and destination travel data, and points of interest (POI) data, and preprocess them to obtain a multidimensional spatiotemporal dataset. The model building module is used to construct a hybrid bidirectional prediction model of supply capacity and travel demand based on the multidimensional spatiotemporal dataset. Based on the physical capacity limit of urban roads and parking resources, and combined with the reasonable modal share of taxis in the traffic structure, the supply capacity is calculated. By integrating deep learning and traffic planning theory, the spatiotemporal network distribution of origin and destination travel is simulated to obtain the initial capacity scale based on travel demand. The optimization module is used to sequentially perform model correction based on multimodal transportation development trends and estimation adjustment based on horizontal city comparison on the initial capacity scale to obtain the optimized recommended capacity scale. The early warning and control module is used to set up a multi-dimensional monitoring indicator system based on the optimized recommended capacity scale, to monitor and evaluate the operation status of the taxi market in real time, and to trigger tiered early warnings and implement corresponding control response measures based on the evaluation results, so as to achieve a dynamic balance between supply and demand of transportation capacity.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the urban vehicle capacity optimization method based on hybrid prediction as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the urban vehicle capacity optimization method based on hybrid prediction as described in any one of claims 1 to 7.

Images