A method, device, equipment and medium for predicting the connection of online car-hailing capacity

Abstract

The application relates to a network car connection capacity prediction method, device, equipment and medium, and relates to the technical field of intelligent traffic and vehicle fleet scheduling. The method comprises the following steps: acquiring multiple order data, order completion data, vehicle real-time position information and external data information of a network car aggregation platform; performing repeated identification and repeated processing on the multiple order data; calculating a historical capacity gap value sequence according to a preset time granularity and a preset geographic fence based on the order data after deduplication, the vehicle real-time position information and the order completion data; constructing a feature matrix based on the historical capacity gap value sequence, the external data information and time period information; inputting the feature matrix into an LSTM prediction model to obtain a capacity gap prediction value set; and determining vehicle scheduling information based on the capacity gap prediction value set, the vehicle real-time position information and a preset scheduling decision algorithm. The application has the effects of accurately predicting the capacity gap of network cars in a future preset time period and reasonably determining vehicle scheduling information.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation and fleet dispatching technology, and in particular to a method, device, equipment and medium for predicting ride-hailing connecting capacity. Background Technology

[0002] In the field of intelligent transportation and fleet dispatching, with the acceleration of urbanization and the increasing travel demands of people, the demand for ride-hailing services is also booming. An efficient ride-hailing dispatching system is of paramount importance for improving urban traffic efficiency, enhancing passenger travel experience, and optimizing the allocation of ride-hailing resources.

[0003] Currently, there are three main approaches to ride-hailing dispatch and capacity prediction. First, statistical methods such as ARIMA / SARIMA, which are characterized by their well-defined structure and simple implementation. Second, machine learning / shallow models, such as SVR, RBF, and GBRT, which can uncover non-linear relationships in data. Third, deep learning / graph network methods, such as CNN, GNN + RNN / LSTM, which show significant advantages in capturing spatiotemporal dependencies.

[0004] However, existing solutions have certain drawbacks. Distribution through aggregation platforms can lead to multiple duplicate records for the same passenger demand, directly causing demand overestimation and severely impacting the accuracy of gap identification and scheduling. Traditional statistical models and some shallow models struggle to capture long-term and short-term dependencies and the impact of sudden events, resulting in a significant drop in prediction accuracy during peak periods, holidays, and other sudden fluctuations. Some complex spatial models struggle to balance accuracy and computational efficiency in engineered real-time inference. Furthermore, many models rely solely on simple historical counts or time variables, lacking multi-scale feature fusion and semantic event representation, failing to effectively capture heterogeneous impacts such as holidays and concerts, and exhibiting insufficient modeling of spatial heterogeneity and station attributes, thus affecting the accuracy of station-level predictions. In addition, some high-precision models have high training and inference costs, making it difficult to support short-granularity rolling online inference. Their model hot updates, concept drift detection, and automatic retraining mechanisms are often incomplete, making low-latency coupling with real-time scheduling systems difficult and affecting the timeliness of decision-making. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method, apparatus, equipment and medium for predicting the connecting capacity of ride-hailing services, in order to solve at least one of the above-mentioned technical problems.

[0006] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: Firstly, this application provides a method for predicting ride-hailing connection capacity, employing the following technical solution: A method for predicting ride-hailing connection capacity includes: The system acquires multiple order data, completion data, real-time vehicle location information, and external data information from ride-hailing aggregation platforms. The order data includes order time, ride time, passenger information, pick-up location, and destination. The external data information includes environmental conditions and public transportation conditions. Duplicate records are identified and processed from multiple order data to obtain deduplicated order data; Based on the deduplicated order data, the real-time vehicle location information, and the completed order data, a historical capacity gap value sequence is calculated according to a preset time granularity and a preset geofence. Based on the historical capacity gap value sequence, the external data information, and the time period information, a feature matrix is ​​constructed. The feature matrix includes time-series features, event features, and periodic features. The feature matrix is ​​input into a pre-trained LSTM prediction model to obtain a set of predicted capacity gap values ​​for a future preset time period. Based on the set of predicted capacity gaps, real-time vehicle location information, and a preset scheduling decision algorithm, vehicle scheduling information is determined, which includes a list of vehicles to be scheduled and the scheduling time.

[0007] The beneficial effects of this invention are as follows: by acquiring multi-source information to provide basic data for subsequent analysis, deduplication of order data improves data quality and ensures the authenticity and reliability of prediction input. Based on the deduplicated data, the historical capacity gap value sequence is calculated and a feature matrix is ​​constructed, which can comprehensively capture the dynamic factors affecting the change of the gap. The LSTM prediction model is used to obtain a high-precision set of predicted capacity gap values ​​for the future preset time period. Based on the set of predicted values, vehicle scheduling information is determined, which can improve the scheduling hit rate and reduce the empty running rate and passenger waiting time.

[0008] Based on the above technical solution, the present invention can be further improved as follows.

[0009] Furthermore, the process of identifying and processing duplicate records from multiple order data to obtain deduplicated order data includes: Based on the order placement time, vehicle usage time, and order number in the order data, a unique identification key is constructed for each order data. For all order data with the same unique identifier key, merge all order data with the same unique identifier key to obtain the target order data; For every two order data that have different unique identification keys, calculate the similarity between the two order data in the passenger information field to obtain the first similarity information; For every two order data with different unique identification keys, calculate the spatial distance similarity between the two order data in the field of pick-up location to obtain the second similarity information; For every two order data that have different unique identification keys, calculate the similarity between the two order data in the destination name or destination latitude and longitude field to obtain third similarity information; For every two order data with different unique identification keys, the first similarity information, the second similarity information, and the third similarity information are weighted and fused according to preset weights to obtain a comprehensive similarity score; For every two order data with different unique identification keys, if the comprehensive similarity score exceeds the preset similarity threshold, the two order data are determined to be duplicate records of the same travel demand. The two duplicate order data are then merged to obtain the target order data. Based on all target order data, determine the deduplicated order data.

[0010] The beneficial effects of adopting the above-mentioned further solution are as follows: by constructing a unique identification key based on the order time, vehicle usage time, and order number, and merging order data with the same key, duplicate orders can be initially identified and processed; for order data with different unique identification keys, similarity is calculated and weighted by considering passenger information, pick-up location, destination name, or latitude and longitude, which can further accurately identify duplicate records of the same travel demand, merge them, and determine the deduplicated order data. This reduces the problem of duplicate records caused by the distribution of aggregation platforms, avoids the overestimation of demand affecting gap identification and scheduling, and improves the authenticity and reliability of the predicted input data.

[0011] Furthermore, the step of calculating a historical capacity gap value sequence based on the deduplicated order data, the real-time vehicle location information, and the completed order data, according to a preset time granularity and a preset geofence, includes: Based on the preset geofence, the physical range of multiple sites is determined; Based on the preset time granularity, the deduplicated order data, real-time vehicle location information, and order completion data are mapped to the corresponding physical range and time window of the depot; For each time window corresponding to each of the aforementioned stations, the total number of orders mapped to the physical range of the station within the time window is counted, and the total number of orders is used as the demand value of the station in the time window; For each time window corresponding to each of the aforementioned stations, based on the real-time vehicle location information and the completed order data, the total number of vehicles whose locations are at or within the geofence and whose status is marked as idle within the time window is counted, and the total number of vehicles is used as the supply value of the station in the time window. For each of the aforementioned stations and each corresponding time window, the historical capacity gap value is determined based on the demand value and the supply value; Based on each historical capacity gap value, a sequence of historical capacity gap values ​​is determined.

[0012] The beneficial effects of adopting the above-mentioned further scheme are as follows: by determining the physical range of the station based on the preset geofence, the data is mapped to the physical range and time window of the station according to the preset time granularity, and the total number of orders and the total number of idle vehicles of each station in each time window are accurately counted, which are used as demand value and supply value respectively. In this way, the historical capacity gap value and sequence are determined, providing an accurate historical data foundation for the subsequent construction of feature matrix and prediction of future capacity gap.

[0013] Furthermore, the construction of a feature matrix based on the historical capacity gap value, the external data information, and the time period information includes: Based on time period information, time period features are extracted and encoded, including hourly segments, weekday segments, weekend identifiers, and holiday identifiers. Based on the historical capacity gap value sequence and the preset sliding window technique, the historical sequence features are determined, and the historical sequence features include historical capacity gap values ​​at multiple time granularities; Based on external data information, external event characteristics are marked, including weather level, public transportation suspension status, and special event identifiers; Based on time period features, historical sequence features, and external event features, a comprehensive feature vector is determined, and the comprehensive feature vector is subjected to one-hot encoding and normalization to obtain the feature matrix.

[0014] The beneficial effects of adopting the above-mentioned further scheme are: by extracting and encoding time period features, determining historical sequence features, and labeling external event features to obtain a feature matrix, it is possible to achieve multi-scale feature fusion and semantic representation of events, comprehensively capture dynamic factors affecting gap changes, improve the ability to capture heterogeneous influences such as holidays and concerts, enhance the modeling of spatial heterogeneity and station attributes, and improve the accuracy of station-level prediction.

[0015] Furthermore, the training method for the pre-trained LSTM prediction model includes: A training sample set is constructed using the historical feature matrix as input and the actual capacity gap value as the prediction target. The loss value is calculated based on the training sample set and the preset mean squared error loss function; Based on the Adam optimizer, the loss value, and the gradient corresponding to the loss value, the parameters of the LSTM prediction model are iteratively optimized to obtain the pre-trained LSTM prediction model. The preset mean squared error loss function is: ; Where θ represents all trainable parameters of the LSTM prediction model, and N represents the number of training samples. This represents the actual capacity gap value for the i-th sample. This represents the predicted capacity gap value for the i-th sample.

[0016] The beneficial effects of adopting the above-mentioned further scheme are: constructing a training sample set with historical feature matrix and actual capacity gap value to provide suitable data for model training; calculating the loss value using the preset mean squared error loss function to measure the gap between the predicted value and the actual value; iteratively optimizing the LSTM prediction model parameters based on Adam optimizer, loss value and its gradient, so that the model continuously adjusts the parameters to achieve better prediction effect, and finally obtains the pre-trained LSTM prediction model.

[0017] Furthermore, the determination of vehicle scheduling information based on the set of predicted capacity gaps, real-time vehicle location information, and a preset scheduling decision algorithm includes: Based on real-time vehicle location and status information, determine the set of available vehicles; Based on the predicted capacity gap set, the set of terminals, the set of available vehicles, and the preset maximum acceptable arrival time threshold, the estimated arrival time of each vehicle to each terminal is calculated, and based on the estimated arrival time of each vehicle to each terminal, the candidate vehicle set corresponding to each terminal is determined. Based on the predicted capacity gap of each station, the estimated arrival time of each vehicle in the candidate vehicle set, the preset demand weight and the preset time weight, calculate the scheduling value score of each station and each vehicle in the corresponding candidate vehicle set. Based on each scheduling value score, preset constraints, and preset objective function, the integer programming model is solved to obtain vehicle scheduling information. The preset optimization conditions are that each vehicle is assigned to a depot and the total number of vehicles assigned to each depot does not exceed the predicted capacity gap of the depot. The preset objective function is a function that aims to maximize the overall scheduling value.

[0018] The beneficial effects of adopting the above-mentioned further scheme are as follows: by determining the set of available vehicles based on the real-time location and status of vehicles, it is possible to accurately locate vehicles that can participate in scheduling; by determining the set of candidate vehicles based on capacity gaps, stations, available vehicles, and the maximum acceptable arrival time threshold, the range of scheduling selection is narrowed; the calculation of scheduling value score comprehensively considers station demand and vehicle arrival time, making scheduling decisions more reasonable; by solving the integer programming model, the overall scheduling value is maximized under the condition of satisfying vehicle allocation constraints, which improves the scheduling hit rate, reduces the empty running rate and passenger waiting time, and provides decision support for the capacity scheduling of ride-hailing services.

[0019] Furthermore, for candidate vehicles belonging to multiple depots, vehicle scheduling information is determined based on a game-theoretic allocation method and each scheduling value score. For candidate vehicles belonging to multiple depots simultaneously, vehicle scheduling information is determined based on a game-theoretic allocation method and various scheduling value scores, including: Based on each scheduling value score, a resource allocation game model is constructed. The resource allocation game model represents vehicles as resources, depots as participants, and scheduling value scores as revenues. Based on the resource allocation game model and the preset initial strategy parameters, determine the resource allocation scheme for the current iteration; The resource utilization efficiency, changes in participant benefits, and overall system performance of the current iteration of the resource allocation scheme are evaluated in real time to obtain the evaluation results; Update the policy parameters of each participant in the current iteration based on the evaluation results; The game is iterated over multiple rounds based on the updated strategy parameters until the change in the resource allocation scheme between two adjacent iterations is less than a preset threshold. Then, the resource allocation scheme corresponding to the current iteration is used as the vehicle scheduling information.

[0020] The beneficial effects of adopting the above-mentioned further solutions are as follows: by acquiring multi-source data and deduplicating it, the authenticity and reliability of the predicted input can be guaranteed, avoiding overestimation of demand due to duplicate recording; based on multi-dimensional accurate calculation of historical capacity gap value sequences and the integration of multi-source information to construct a feature matrix, dynamic factors affecting gap changes can be comprehensively captured; using the LSTM prediction model for prediction can effectively capture long-term and short-term dependencies, achieving high-precision capacity gap prediction; on this basis, using a game-theoretic allocation method to schedule candidate vehicles belonging to multiple stations can construct a resource allocation game model, continuously optimizing the resource allocation scheme through iterative game theory, reducing resource overlap scheduling, improving scheduling hit rate and reducing empty running rate.

[0021] Secondly, this application provides a ride-hailing service connection capacity prediction device, which adopts the following technical solution: A ride-hailing service connection capacity prediction device includes: The data acquisition module is used to acquire multiple order data, order completion data, real-time vehicle location information, and external data information from the ride-hailing aggregation platform. The order data includes order time, ride time, passenger information, pick-up location, and destination. The external data information includes environmental status and public transportation status. The data cleaning and deduplication module is used to identify and process duplicate records from multiple order data to obtain deduplicated order data. The feature engineering module is used to calculate the historical capacity gap value based on the deduplicated order data, the real-time vehicle location information, and the completed order data, according to a preset time granularity and a preset geofence, and to construct a feature matrix based on the historical capacity gap value, the external data information, and the time period information. The feature matrix includes time-series features, event features, and periodic features. The LSTM prediction module is used to input the feature matrix into a pre-trained LSTM prediction model to obtain the predicted capacity gap for a future preset period. The scheduling linkage module is used to determine vehicle scheduling information based on the predicted capacity gap, real-time vehicle location information, and a preset scheduling decision algorithm. The vehicle scheduling information includes a list of vehicles to be scheduled and the scheduling time.

[0022] Thirdly, this application provides an electronic device that adopts the following technical solution: An electronic device includes a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and execute the ride-hailing connection capacity prediction method according to any one of the first aspects.

[0023] Fourthly, this application provides a computer-readable storage medium, which adopts the following technical solution: A computer-readable storage medium storing a computer program capable of being loaded by a processor and executing the ride-hailing connection capacity prediction method as described in any of the first aspects.

[0024] Additional aspects and advantages of this application will be set forth in part in the description which follows, and will become apparent from the description or may be learned by practice of this application. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating a method for predicting ride-hailing connection capacity according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of a ride-hailing connection capacity prediction device provided in one embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of an electronic device provided in one embodiment of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0027] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0028] This application provides a method for predicting ride-hailing connection capacity. This method can be executed by an electronic device, which can be a server or a mobile terminal device. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services. The mobile terminal device can be a laptop, a desktop computer, etc., but is not limited to these.

[0029] like Figure 1 As shown, a method for predicting ride-hailing connection capacity mainly includes: S1, acquire multiple order data, order completion data, real-time vehicle location information and external data information from the ride-hailing aggregation platform. The order data includes order time, ride time, passenger information, pick-up location and destination. The external data information includes environmental status and public transportation status. In this embodiment of the application, multiple order data, order completion data, real-time vehicle location information and external data information can be collected using data acquisition devices, platform API interfaces or log receivers.

[0030] Data acquisition equipment can be specialized data collectors. The platform's API interface can seamlessly connect with ride-hailing aggregation platforms to acquire the necessary data in real time; log receivers can collect and organize various log data generated by the platform. The purpose of these data acquisition devices is to obtain multiple order data, completion data, real-time vehicle location information, and external data information from the ride-hailing aggregation platform. Order data includes order time, ride time, passenger information, pick-up location, and destination; external data information includes environmental conditions and public transportation status.

[0031] S2, perform duplicate record identification and duplicate processing on multiple order data to obtain the deduplicated order data; In this embodiment of the application, specifically, S2 includes the following sub-steps: S21, Based on the order placement time, vehicle usage time and order number in the order data, construct a unique identification key for each order data; S22, For all order data with the same unique identification key, merge all order data with the same unique identification key to obtain the target order data; S23, for every two order data with different unique identification keys, calculate the similarity between the two order data in the passenger information field to obtain the first similarity information; S24, For every two order data with different unique identification keys, calculate the spatial distance similarity of the two order data in the field of pick-up location to obtain the second similarity information; S25, for every two order data with different unique identification keys, calculate the similarity between the two order data in the destination name or destination latitude and longitude field to obtain third similarity information; S26. For every two order data with different unique identification keys, the first similarity information, the second similarity information and the third similarity information are weighted and fused according to preset weights to obtain a comprehensive similarity score. S27. For every two order data with different unique identification keys, if the comprehensive similarity score exceeds the preset similarity threshold, the two order data are determined to be duplicate records of the same travel demand. The two duplicate order data are merged to obtain the target order data. S28, Based on all target order data, determine the deduplicated order data.

[0032] In this embodiment, uniqueness identification requires constructing a unique identification key for each order based on the order placement time, ride time, and order number in the order data. This identification key acts like an "ID number" for the order, quickly distinguishing different orders. Similarity comparison calculates the similarity of order data with different unique identification keys across fields such as passenger information, pick-up location, and destination. For example, similarity can be calculated by comparing passenger phone numbers, pick-up location latitude and longitude, and destination name or latitude and longitude. These similarity scores are then weighted and fused according to preset weights to obtain a comprehensive similarity score. If the comprehensive similarity score exceeds a preset similarity threshold, the two order data are determined to be duplicate records of the same travel demand, and they are merged to obtain the target order data. This effectively avoids the problem of overestimating demand due to duplicate records and improves the accuracy of subsequent data analysis.

[0033] S3, based on the deduplicated order data, the real-time vehicle location information, and the completed order data, calculate the historical capacity gap value sequence according to the preset time granularity and the preset geofence, and construct a feature matrix based on the historical capacity gap value sequence, the external data information, and the time period information. The feature matrix includes time-series features, event features, and periodic features. In this embodiment of the application, the step of calculating a historical capacity gap value sequence based on the deduplicated order data, the real-time vehicle location information, and the completed order data, according to a preset time granularity and a preset geofence, includes: Based on the preset geofence, the physical range of multiple sites is determined; Based on the preset time granularity, the deduplicated order data, real-time vehicle location information, and order completion data are mapped to the corresponding physical range and time window of the depot; For each time window corresponding to each of the aforementioned stations, the total number of orders mapped to the physical range of the station within the time window is counted, and the total number of orders is used as the demand value of the station in the time window; For each time window corresponding to each of the aforementioned stations, based on the real-time vehicle location information and the completed order data, the total number of vehicles whose locations are at or within the geofence and whose status is marked as idle within the time window is counted, and the total number of vehicles is used as the supply value of the station in the time window. For each of the aforementioned stations and each corresponding time window, the historical capacity gap value is determined based on the demand value and the supply value; Based on each historical capacity gap value, a sequence of historical capacity gap values ​​is determined.

[0034] In this embodiment, the physical boundaries of multiple stations can be defined using technologies such as Geographic Information Systems (GIS). The historical capacity gap can be determined by calculating the difference between demand and supply.

[0035] In this embodiment of the application, the step of constructing a feature matrix based on the historical capacity gap value, the external data information, and the time period information includes: Based on time period information, time period features are extracted and encoded, including hourly segments, weekday segments, weekend identifiers, and holiday identifiers. Based on the historical capacity gap value sequence and the preset sliding window technique, the historical sequence features are determined, and the historical sequence features include historical capacity gap values ​​at multiple time granularities; Based on external data information, external event characteristics are marked, including weather level, public transportation suspension status, and special event identifiers; Based on time period features, historical sequence features, and external event features, a comprehensive feature vector is determined, and the comprehensive feature vector is subjected to one-hot encoding and normalization to obtain the feature matrix.

[0036] The constructed feature matrix can more comprehensively reflect relevant information about ride-hailing capacity, providing strong support for subsequent predictions.

[0037] S4, input the feature matrix into the pre-trained LSTM prediction model to obtain a set of predicted capacity gap values ​​for a future preset period; In this embodiment of the application, the training method for the pre-trained LSTM prediction model includes: A training sample set is constructed using the historical feature matrix as input and the actual capacity gap value as the prediction target. The loss value is calculated based on the training sample set and the preset mean squared error loss function; Based on the Adam optimizer, the loss value, and the gradient corresponding to the loss value, the parameters of the LSTM prediction model are iteratively optimized to obtain the pre-trained LSTM prediction model. The preset mean squared error loss function is: ; Where θ represents all trainable parameters of the LSTM prediction model, and N represents the number of training samples. This represents the actual capacity gap value for the i-th sample. This represents the predicted capacity gap value for the i-th sample.

[0038] In the above implementation, the training of the pre-trained LSTM prediction model requires a historical feature matrix as input and the actual capacity gap value as the prediction target to construct a training sample set. The loss value is calculated using a preset mean squared error loss function, and then the LSTM prediction model parameters are iteratively optimized using the Adam optimizer, the loss value, and the corresponding gradient. The mean squared error loss function measures the degree of error between the predicted and actual values, while the Adam optimizer adaptively adjusts the model parameters, improving the training efficiency and accuracy of the model. After multiple iterations of optimization, the pre-trained LSTM prediction model is obtained. Inputting the feature matrix into this model yields a set of predicted capacity gap values ​​for a preset future time period. Due to its powerful time-series memory function, the LSTM model can better capture the changing trends of capacity demand, especially during peak periods and holidays, thus improving prediction accuracy.

[0039] The pre-trained LSTM prediction model includes: An input layer is used to receive a feature matrix, which contains multivariate features of the time step sequence; An LSTM layer consists of at least one LSTM unit, and each LSTM unit performs the following computation at time step t: Forgotten Gate: ; Input gates and candidate cells: ; ; Cell status update: ; Output gates and hidden states: ; ; Where σ represents the sigmoid activation function, tanh represents the hyperbolic tangent activation function, and ⊙ represents element-wise multiplication. The input feature vector represents time step t; This represents the hidden state at time step t-1; This represents the cell state at time step t-1; , , , These represent the weight matrices corresponding to the forget gate, input gate, candidate cell state, and output gate, respectively. , , , These represent the bias vectors corresponding to the forget gate, input gate, candidate cell state, and output gate, respectively. A fully connected layer is used to store the hidden state of the last time step of the LSTM layer. Perform a linear transformation; The output layer is used to output the predicted capacity gap for a future preset time period and the corresponding confidence level estimate.

[0040] The parameter update rules for the Adam optimizer include: Calculate the first moment estimate of the gradient: ; Calculate the second moment estimate of the gradient: ; Bias correction for the first-order moment estimate: ; Bias correction for the second-order moment estimate: ; Update model parameters: ; Where t represents the number of training iterations. This represents the gradient of the loss function at time step t. This represents the first moment estimate of the gradient. This represents the second moment estimate of the gradient.

[0041] The statistical formulas used in feature engineering include: (1) Date definition: ; (2) Definition of month variable: ; (3) Definition of weekday variable: ; (4) Range of values ​​for the time period: (5) Predict the arrival of ride-hailing orders to fill the gap. This represents the gap order to be predicted at time t. , ,..., As a predictive feature, Kh represents the number of preceding time periods.

[0042] (6) Orders placed on previous days: This represents the order gap at time t on day d. These are the order gaps from the same time period in the previous k days.

[0043] (7) Previous week's orders: This represents the order gap at time t on day d. These are the order gaps from the same time period in the previous k weeks.

[0044] S5. Based on the set of predicted capacity gaps, real-time vehicle location information, and a preset scheduling decision algorithm, vehicle scheduling information is determined, including a list of vehicles to be scheduled and scheduling time.

[0045] In this embodiment of the application, the step of determining vehicle dispatch information based on the set of predicted capacity gaps, real-time vehicle location information, and a preset dispatch decision algorithm includes: Based on real-time vehicle location and status information, determine the set of available vehicles; Based on the predicted capacity gap set, the set of terminals, the set of available vehicles, and the preset maximum acceptable arrival time threshold, the estimated arrival time of each vehicle to each terminal is calculated, and based on the estimated arrival time of each vehicle to each terminal, the candidate vehicle set corresponding to each terminal is determined. Based on the predicted capacity gap of each station, the estimated arrival time of each vehicle in the candidate vehicle set, the preset demand weight and the preset time weight, calculate the scheduling value score of each station and each vehicle in the corresponding candidate vehicle set. Based on each scheduling value score, preset constraints, and preset objective function, the integer programming model is solved to obtain vehicle scheduling information. The preset optimization conditions are that each vehicle is assigned to a depot and the total number of vehicles assigned to each depot does not exceed the predicted capacity gap of the depot. The preset objective function is a function that aims to maximize the overall scheduling value.

[0046] In this embodiment, the set of available vehicles is first determined based on the vehicle's real-time location information and vehicle status information. The vehicle's location information can be obtained in real time through its positioning system, and its availability can be determined in conjunction with the vehicle's operational status.

[0047] Then, based on the predicted capacity gap set, the set of depots, the set of available vehicles, and the preset maximum acceptable arrival time threshold, the estimated arrival time of each vehicle to each depot is calculated. By breaking down the vehicle's journey to the depot into multiple segments, and combining real-time traffic information and historical driving data, the travel time of each segment is estimated, and finally, these are summed to obtain the estimated arrival time. Based on the estimated arrival time of each vehicle to each depot, a candidate vehicle set corresponding to each depot is determined.

[0048] Next, based on the predicted capacity gap for each depot, the estimated arrival time of each vehicle in the candidate vehicle set, the preset demand weight, and the preset time weight, the scheduling value score for each depot and each vehicle in the corresponding candidate vehicle set is calculated. The scheduling value score comprehensively considers the demand intensity of the depot and the time efficiency of the vehicles, enabling a more reasonable assessment of vehicle scheduling priorities.

[0049] Finally, based on each scheduling value score, preset constraints, and preset objective function, the integer programming model is solved to obtain vehicle scheduling information, which includes a list of scheduled vehicles and scheduling time.

[0050] The preset constraints ensure that each vehicle is assigned to a station and that the total number of vehicles assigned to each station does not exceed the predicted capacity gap of that station. The preset objective function is a function that aims to maximize the overall scheduling value, which can ensure the rational allocation and efficient utilization of resources.

[0051] In this embodiment, the vehicle's "value score" for the depot is calculated. The value score model consists of a linear combination of two main components: a depot demand intensity term and a time efficiency term, with a demand weight of 0.7 and a time weight of 0.3, as shown in the example below: Factors affecting the intensity of demand at the station: ; in It is the predicted gap number for station s in the current time window. It is the largest gap value among all stations. This is the gap weight, usually set to 0.7; Time efficiency factors: ; in This is the estimated arrival time of vehicle v at depot s. This represents the maximum arrival time of the candidate vehicles to each depot. This is the time weight, which is usually set to 0.3.

[0052] Optimal allocation decision: The integer programming optimal allocation method based on utility function expresses the allocation in the form of integer programming, which facilitates the explanation of the solution objective and constraints. The final output includes the scheduling range, the number of vehicles to be scheduled, and the list of vehicles to be scheduled.

[0053] .

[0054] As a further optional implementation method in the embodiments of this application, it also includes: For candidate vehicles belonging to multiple depots, vehicle scheduling information is determined based on a game-theoretic allocation method and each scheduling value score. For candidate vehicles belonging to multiple depots simultaneously, vehicle scheduling information is determined based on a game-theoretic allocation method and various scheduling value scores, including: Based on each scheduling value score, a resource allocation game model is constructed. The resource allocation game model represents vehicles as resources, depots as participants, and scheduling value scores as revenues. Based on the resource allocation game model and the preset initial strategy parameters, determine the resource allocation scheme for the current iteration; The resource utilization efficiency, changes in participant benefits, and overall system performance of the current iteration of the resource allocation scheme are evaluated in real time to obtain the evaluation results; Update the policy parameters of each participant in the current iteration based on the evaluation results; The game is iterated over multiple rounds based on the updated strategy parameters until the change in the resource allocation scheme between two adjacent iterations is less than a preset threshold. Then, the resource allocation scheme corresponding to the current iteration is used as the vehicle scheduling information.

[0055] This game-theoretic allocation method can more rationally allocate candidate vehicles belonging to multiple depots, avoiding resource waste and conflicts, and improving the overall scheduling efficiency.

[0056] This method provides foundational data for subsequent analysis by acquiring information from multiple sources. It improves data quality by deduplicating order data and ensures the authenticity and reliability of predictive inputs. Based on the deduplicated data, it calculates the historical capacity gap value sequence and constructs a feature matrix, which can comprehensively capture the dynamic factors affecting gap changes. It uses an LSTM prediction model to obtain a high-precision set of predicted capacity gap values ​​for a future preset time period. Based on the set of predicted values, it determines vehicle scheduling information, which can improve scheduling hit rate and reduce empty running rate and passenger waiting time.

[0057] Figure 2 This is a schematic diagram of the structure of a ride-hailing connection capacity prediction device 200.

[0058] like Figure 2 As shown, a ride-hailing connection capacity prediction device 200 mainly includes: The data acquisition module 201 is used to acquire multiple order data, order completion data, real-time vehicle location information and external data information from the ride-hailing aggregation platform. The order data includes order time, ride time, passenger information, pick-up location and destination. The external data information includes environmental status and public transportation status. The data cleaning and deduplication module 202 is used to identify and process duplicate records in multiple order data to obtain deduplicated order data. The feature engineering module 203 is used to calculate the historical capacity gap value based on the deduplicated order data, the real-time vehicle location information and the completed order data, according to a preset time granularity and a preset geofence, and to construct a feature matrix based on the historical capacity gap value, the external data information and time period information. The feature matrix includes time-series features, event features and periodic features. The LSTM prediction module 204 is used to input the feature matrix into a pre-trained LSTM prediction model to obtain the predicted value of the capacity gap for a future preset period. The scheduling linkage module 205 is used to determine vehicle scheduling information based on the predicted value of the transportation capacity gap, the real-time location information of vehicles and the preset scheduling decision algorithm. The vehicle scheduling information includes a list of vehicles to be scheduled and the scheduling time.

[0059] In one example, the module in any of the above devices may be one or more integrated circuits configured to implement the above methods, such as one or more application-specific integrated circuits (ASICs), or one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs), or a combination of at least two of these integrated circuit forms.

[0060] For example, when modules in a device can be implemented via a processing element scheduler, the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processor capable of calling programs. Alternatively, these modules can be integrated together as a system-on-a-chip (SOC).

[0061] In this application, various objects such as messages / information / devices / network elements / systems / apparatus / actions / operations / processes / concepts may be named. It is understood that these specific names do not constitute a limitation on the relevant objects. The names may be changed depending on the scenario, context, or usage habits. The understanding of the technical meaning of the technical terms in this application should be mainly determined from their functions and technical effects embodied / performed in the technical solution.

[0062] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0063] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0064] Figure 3 This is a structural block diagram of an electronic device 300 according to an embodiment of this application.

[0065] like Figure 3As shown, the electronic device 300 includes a processor 301 and a memory 302, and may further include one or more of an information input / output (I / O) interface 303, a communication component 304, and a communication bus 305.

[0066] The processor 301 controls the overall operation of the electronic device 300 to complete all or part of the steps in the aforementioned method for predicting ride-hailing connection capacity. The memory 302 stores various types of data to support the operation of the electronic device 300. This data may include, for example, instructions for any application or method operating on the electronic device 300, as well as application-related data. The memory 302 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as one or more of Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0067] I / O interface 303 provides an interface between processor 301 and other interface modules, such as keyboards, mice, and buttons. These buttons can be virtual or physical. Communication component 304 is used to test wired or wireless communication between electronic device 300 and other devices. Wireless communication includes Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 304 may include a Wi-Fi component, a Bluetooth component, and an NFC component.

[0068] The communication bus 305 may include a path for transmitting information between the aforementioned components. The communication bus 305 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The communication bus 305 may be divided into an address bus, a data bus, a control bus, etc.

[0069] The electronic device 300 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to execute the ride-hailing service capacity prediction method given in the above embodiments.

[0070] The following describes the computer-readable storage medium provided in the embodiments of this application. The computer-readable storage medium described below can be referred to in correspondence with the ride-hailing connection capacity prediction method described above.

[0071] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for predicting ride-hailing connection capacity.

[0072] The computer-readable storage medium may include 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.

[0073] The terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0074] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the foregoing application concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions claimed in this application.

Claims

1. A method for predicting ride-hailing connection capacity, characterized in that, include: The system acquires multiple order data, completion data, real-time vehicle location information, and external data information from ride-hailing aggregation platforms. The order data includes order time, ride time, passenger information, pick-up location, and destination. The external data information includes environmental conditions and public transportation conditions. Duplicate records are identified and processed from multiple order data to obtain deduplicated order data; Based on the deduplicated order data, the real-time vehicle location information, and the completed order data, a historical capacity gap value sequence is calculated according to a preset time granularity and a preset geofence. Based on the historical capacity gap value sequence, the external data information, and the time period information, a feature matrix is ​​constructed. The feature matrix includes time-series features, event features, and periodic features. The feature matrix is ​​input into a pre-trained LSTM prediction model to obtain a set of predicted capacity gap values ​​for a future preset time period. Based on the set of predicted capacity gaps, real-time vehicle location information, and a preset scheduling decision algorithm, vehicle scheduling information is determined, which includes a list of vehicles to be scheduled and the scheduling time.

2. The method for predicting ride-hailing connection capacity according to claim 1, characterized in that, The process of identifying and processing duplicate records from multiple order data to obtain deduplicated order data includes: Based on the order placement time, vehicle usage time, and order number in the order data, a unique identification key is constructed for each order data. For all order data with the same unique identifier key, merge all order data with the same unique identifier key to obtain the target order data; For every two order data that have different unique identification keys, calculate the similarity between the two order data in the passenger information field to obtain the first similarity information; For every two order data with different unique identification keys, calculate the spatial distance similarity between the two order data in the field of pick-up location to obtain the second similarity information; For every two order data that have different unique identification keys, calculate the similarity between the two order data in the destination name or destination latitude and longitude field to obtain third similarity information; For every two order data with different unique identification keys, the first similarity information, the second similarity information, and the third similarity information are weighted and fused according to preset weights to obtain a comprehensive similarity score; For every two order data with different unique identification keys, if the comprehensive similarity score exceeds the preset similarity threshold, the two order data are determined to be duplicate records of the same travel demand. The two duplicate order data are then merged to obtain the target order data. Based on all target order data, determine the deduplicated order data.

3. The method for predicting ride-hailing connection capacity according to claim 2, characterized in that, The step of calculating a historical capacity gap value sequence based on the deduplicated order data, the real-time vehicle location information, and the completed order data, according to a preset time granularity and a preset geofence, includes: Based on the preset geofence, the physical range of multiple sites is determined; Based on the preset time granularity, the deduplicated order data, real-time vehicle location information, and order completion data are mapped to the corresponding physical range and time window of the depot; For each time window corresponding to each of the aforementioned stations, the total number of orders mapped to the physical range of the station within the time window is counted, and the total number of orders is used as the demand value of the station in the time window; For each time window corresponding to each of the aforementioned stations, based on the real-time vehicle location information and the order completion data, the total number of vehicles whose location is at or within the geofence and whose status is marked as idle within the time window is counted, and the total number of vehicles is used as the supply value of the station in the time window. For each of the aforementioned stations and each corresponding time window, the historical capacity gap value is determined based on the demand value and the supply value; Based on each historical capacity gap value, a sequence of historical capacity gap values ​​is determined.

4. The method for predicting ride-hailing connection capacity according to claim 2, characterized in that, The construction of a feature matrix based on the historical capacity gap value, the external data information, and the time period information includes: Based on time period information, time period features are extracted and encoded, including hourly segments, weekday segments, weekend identifiers, and holiday identifiers. Based on the historical capacity gap value sequence and the preset sliding window technique, the historical sequence features are determined, and the historical sequence features include historical capacity gap values ​​at multiple time granularities; Based on external data information, external event characteristics are marked, including weather level, public transportation suspension status, and special event identifiers; Based on time period features, historical sequence features, and external event features, a comprehensive feature vector is determined, and the comprehensive feature vector is subjected to one-hot encoding and normalization to obtain the feature matrix.

5. The method for predicting ride-hailing connection capacity according to claim 1, characterized in that, The training method for the pre-trained LSTM prediction model includes: A training sample set is constructed using the historical feature matrix as input and the actual capacity gap value as the prediction target. The loss value is calculated based on the training sample set and the preset mean squared error loss function; Based on the Adam optimizer, the loss value, and the gradient corresponding to the loss value, the parameters of the LSTM prediction model are iteratively optimized to obtain the pre-trained LSTM prediction model. The preset mean squared error loss function is: ； Where θ represents all trainable parameters of the LSTM prediction model, and N represents the number of training samples. This represents the actual capacity gap value for the i-th sample. This represents the predicted capacity gap value for the i-th sample.

6. The method for predicting ride-hailing connection capacity according to claim 3, characterized in that, The method for determining vehicle dispatch information based on the set of predicted capacity gaps, real-time vehicle location information, and a pre-defined dispatch decision algorithm includes: Based on real-time vehicle location and status information, determine the set of available vehicles; Based on the predicted capacity gap set, the set of terminals, the set of available vehicles, and the preset maximum acceptable arrival time threshold, the estimated arrival time of each vehicle to each terminal is calculated, and based on the estimated arrival time of each vehicle to each terminal, the candidate vehicle set corresponding to each terminal is determined. Based on the predicted capacity gap of each station, the estimated arrival time of each vehicle in the candidate vehicle set, the preset demand weight and the preset time weight, calculate the scheduling value score of each station and each vehicle in the corresponding candidate vehicle set. Based on each scheduling value score, preset constraints, and preset objective function, the integer programming model is solved to obtain vehicle scheduling information. The preset optimization conditions are that each vehicle is assigned to a depot and the total number of vehicles assigned to each depot does not exceed the predicted capacity gap of the depot. The preset objective function is a function that aims to maximize the overall scheduling value.

7. The method for predicting ride-hailing connection capacity according to claim 6, characterized in that, Also includes: For candidate vehicles belonging to multiple depots, vehicle scheduling information is determined based on a game-theoretic allocation method and each scheduling value score. For candidate vehicles belonging to multiple depots simultaneously, vehicle scheduling information is determined based on a game-theoretic allocation method and various scheduling value scores, including: Based on each scheduling value score, a resource allocation game model is constructed. The resource allocation game model represents vehicles as resources, depots as participants, and scheduling value scores as revenues. Based on the resource allocation game model and the preset initial strategy parameters, determine the resource allocation scheme for the current iteration; The resource utilization efficiency, changes in participant benefits, and overall system performance of the current iteration of the resource allocation scheme are evaluated in real time to obtain the evaluation results; Update the policy parameters of each participant in the current iteration based on the evaluation results; The game is iterated over multiple rounds based on the updated strategy parameters until the change in the resource allocation scheme between two adjacent iterations is less than a preset threshold. Then, the resource allocation scheme corresponding to the current iteration is used as the vehicle scheduling information.

8. A device for predicting the connecting capacity of ride-hailing services, characterized in that, include: The data acquisition module is used to acquire multiple order data, order completion data, real-time vehicle location information, and external data information from the ride-hailing aggregation platform. The order data includes order time, ride time, passenger information, pick-up location, and destination. The external data information includes environmental status and public transportation status. The data cleaning and deduplication module is used to identify and process duplicate records from multiple order data to obtain deduplicated order data. The feature engineering module is used to calculate the historical capacity gap value based on the deduplicated order data, the real-time vehicle location information, and the completed order data, according to a preset time granularity and a preset geofence, and to construct a feature matrix based on the historical capacity gap value, the external data information, and the time period information. The feature matrix includes time-series features, event features, and periodic features. The LSTM prediction module is used to input the feature matrix into a pre-trained LSTM prediction model to obtain the predicted capacity gap for a future preset period. The scheduling linkage module is used to determine vehicle scheduling information based on the predicted capacity gap, real-time vehicle location information, and a preset scheduling decision algorithm. The vehicle scheduling information includes a list of vehicles to be scheduled and the scheduling time.

9. An electronic device, characterized in that, Includes a processor, which is coupled to a memory; The processor is configured to execute a computer program stored in the memory to cause the electronic device to perform the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Includes a computer program or instructions that, when run on a computer, cause the computer to perform the method as described in any one of claims 1-7.

Images