Airport public transport pre-allocation processing system and method based on big data model

Abstract

This invention belongs to the field of airport public transportation allocation technology, and discloses an airport public transportation pre-allocation processing system and method based on a big data model. The method includes: acquiring historical data; using big data processing technology and machine learning and deep learning models to infer passenger demand and public transportation usage from historical data, generating a data projection model; inputting real-time weather, time, and passenger data into the data projection model, employing intelligent optimization algorithms, and dynamically adjusting public transportation resource allocation in conjunction with real-time traffic conditions; monitoring the public transportation resource allocation results in real time, continuously optimizing and dynamically adjusting the results using a feedback mechanism, and pushing the latest adjusted results to downstream third-party systems. This invention utilizes real-time data analysis and intelligent optimization algorithms to achieve a more efficient traffic scheduling scheme, and has significant practical implications and broad application prospects.

Description

Technical Field

[0001] This invention belongs to the field of airport public transportation allocation technology, and particularly relates to an airport public transportation pre-allocation processing system and method based on big data models. Background Technology

[0002] With the rapid development of the global air transport industry, airport passenger traffic is constantly increasing, posing significant challenges to the scheduling and allocation of public transportation systems. As urban integrated transportation hubs, the rational allocation of public transportation systems at airports is crucial for improving passenger travel experience, reducing traffic congestion, and increasing operational efficiency. However, existing airport public transportation allocation systems have several shortcomings, mainly in the following aspects:

[0003] Information lag: Currently, most airports rely on fixed timetables and manual scheduling for public transportation, making it difficult to respond in real time to flight dynamics, passenger flow changes, and emergencies, resulting in unreasonable resource allocation.

[0004] Lack of intelligent optimization: Existing allocation methods usually rely on historical experience or static rules, failing to fully utilize big data analysis and machine learning technologies, resulting in low prediction accuracy and inflexible scheduling schemes.

[0005] Poor passenger experience: Due to the uneven distribution of transportation resources, passengers may face problems such as long waiting times, overloaded transportation, or waste of resources at the airport, which affects travel satisfaction.

[0006] Low resource utilization: Public transportation around the airport (such as taxis, buses, subways, etc.) has not been dynamically matched with actual demand, resulting in either resource surplus or shortage, which reduces overall operational efficiency.

[0007] In recent years, the development of big data technology and artificial intelligence has provided new possibilities for optimizing airport public transportation scheduling. Through in-depth analysis of flight information, passenger flow, vehicle status, and historical data, intelligent prediction and dynamic allocation of public transportation resources can be achieved, improving the operational efficiency of airport transportation systems, reducing passenger waiting times, and enhancing the overall travel experience. Summary of the Invention

[0008] To overcome the problems existing in related technologies, this invention discloses an airport public transportation pre-allocation processing system and method based on a big data model. The purpose of this invention is to solve problems such as information lag, lack of intelligent optimization, poor passenger experience, and low resource utilization in existing technologies.

[0009] The technical solution is as follows: A pre-allocation method for airport public transportation based on a big data model, comprising the following steps:

[0010] S1 acquires historical data from multiple sources, including flight information, passenger flow, weather conditions, transportation status, passenger arrival and departure information, and passenger travel intentions.

[0011] S2, based on big data processing technology, uses machine learning and deep learning models to infer passenger demand and public transportation usage from historical data and generate data projection models.

[0012] S3 inputs real-time weather, time, and passenger data into the data extrapolation model, and uses intelligent optimization algorithms to dynamically adjust the allocation of public transportation resources in conjunction with real-time traffic conditions;

[0013] S4 monitors the allocation of public transportation resources in real time, continuously optimizes and dynamically adjusts the results based on feedback mechanisms, and pushes the latest corrected results to downstream third-party systems.

[0014] In step S1, the historical data of multi-source data, including flight information, passenger flow, weather conditions, transportation status, passenger arrival and departure information, and passenger travel intentions, are obtained, including: data collection, data preprocessing, data clustering, and data distribution.

[0015] Data collection includes obtaining flight information, passenger flow, passenger arrival and departure information, passenger travel intentions, weather conditions in the city where the airport is located, and traffic conditions in the city where the airport is located by connecting with platforms such as production and operation, passenger service, airport environmental management, and local traffic management.

[0016] Data preprocessing involves performing preliminary processing on the collected data to filter out useless and erroneous data.

[0017] Data clustering is the process of grouping data according to the time period and the related information of the passengers involved.

[0018] Data distribution involves distributing processed and clustered data to downstream processes.

[0019] In step S2, based on big data processing technology, a large amount of historical data is collected, cleaned, and analyzed to eliminate abnormal interference data. Finally, a set of historical data suitable for machine learning and deep learning is obtained. Then, machine learning and deep learning models are used to infer passenger demand and public transportation usage in the historical data and generate a data inference model. This includes: historical data analysis, model inference and correction.

[0020] Historical data analysis includes: using machine learning and deep learning models to analyze all historical data and predict airport traffic usage under different times, weather conditions, and passenger numbers.

[0021] Set the time-influencing factor to the current time. , The influencing factors change over time, as expressed by:

[0022] (1)

[0023] In the formula, As a time-related factor, It is half the difference between the highest and lowest number of people on that day. Indicates average foot traffic;

[0024] Based on historical data, the formula evolved as follows:

[0025] (2)

[0026] In the formula, For vehicle usage, The weights of each influencing factor (e.g., for a certain 4F-level airport in China, the weight of its influence is...) It is 0.42. It is 0.12. It is 0.09. It is 0.11. It is 0.43. (0.29) Weather type For humidity, For temperature, This is the error value. For passenger flow, Indicate whether it is a holiday;

[0027] Model extrapolation and correction include: analyzing the predicted usage of public transportation in historical data and comparing it with the actual situation at the time; substituting the actual weather and time data into formula (2); judging whether the result Y is similar to the extrapolated result; if similar, no adjustment is made; if not similar, adjustments are made. The value is adjusted so that the deduction result is close to the final actual result; the above steps are repeated. When the deduction result is always close to the actual result, the deduction model can be considered successful, and a data deduction model is generated. This model implements the above formula by writing code and includes the implementation logic of the above formula and the historical data set used for deduction.

[0028] Furthermore, the generated data extrapolation model includes:

[0029] The collected data undergoes preliminary processing, starting with the removal of duplicate and null values ​​using a computer.

[0030] Feature extraction is performed on the processed data;

[0031] Build a dataset by constructing a mapping set between the processed data and public transportation usage;

[0032] The training model uses a large-scale combined model of a Long Short-Term Memory (LSTM) network and an attention-based Transformer model as the data inference model to handle short-term time dependence and long-term trends respectively, and obtain prediction results.

[0033] Training and optimization involve inputting all the constructed datasets into the LSTM and Transformer models and outputting the corresponding results.

[0034] Furthermore, the collected data undergoes preliminary processing. First, a computer-based approach is used to initially remove duplicate and null values. This includes: for data retrieved from Excel, a Python script is written, incorporating the pandas library, to read all data from the Excel file using `pd.read_excel`, iterating through each data entry, deleting the current entry, and saving all remaining data; for data retrieved from a database, a corresponding SQL script is written to determine if the key field `columns` is not null or an empty string, using an encoded method to check for 'columns is not null or columns !=', and finally retrieving all the data. Then, the preliminarily cleared data is manually reviewed to process and correct any abnormal data.

[0035] The feature extraction of the processed data includes: removing useless feature values, then marking and extracting key features, such as whether it is currently a peak time, the coverage area of ​​the peak time, the impact of the corresponding weather on the transportation, and what kind of transportation the passenger will take next.

[0036] Furthermore, LSTM is used to process short-term time series patterns, including:

[0037] Forgotten Gate The formula for determining how much past information is forgotten is:

[0038] ;

[0039] In the formula, The activation function in LSTM. This represents the weight of the current input, with a value between 0 and 1, depending on the input content. The input data includes current hourly flight information and weather conditions; The weights input at the previous time step. This refers to the data entered at the previous moment, including flight information and weather conditions. This is an offset value; the final calculation result is obtained by manually intervening in the numerical value. For the current moment, The previous moment;

[0040] Input gate This determines how the currently input information should be stored in the cell state, and remembers the impact of severe weather on traffic. The formula is:

[0041] ;

[0042] In the formula, This is the weight of the current input gate. Here, the importance of the input is judged manually based on the input content and adjusted manually. The value is between 0 and 1. The weights input to the previous input gate, This is the input gate offset;

[0043] Cell state It is responsible for storing long-term memory information and remembering past peak patterns. The formula is:

[0044] ;

[0045] In the formula, This represents the numerical value of the cell state data from the previous moment. It is the hyperbolic tangent function in mathematics;

[0046] Output gate The formula determines how much information about the current cell state to output, and only outputs information that affects future passenger flow.

[0047] ;

[0048] In the formula, The weights of the current output gate. The weight of the output gate at the previous time step. This is the output gate offset;

[0049] Final result , representing the public transportation forecast for the next moment, is calculated using the following formula:

[0050] ;

[0051] In the formula, This is the value obtained from the cell state formula at the current moment.

[0052] Furthermore, the Transformer is used to capture long-term trends and global patterns, combined with LSTM to learn short-term passenger flow changes, as shown in the formula:

[0053] ;

[0054] In the formula, This is the result of the calculation formula used in this experiment; This is a normalization function that transforms numerical values ​​into a probability distribution. For matrix transpose, Given a matrix, its transpose is ANDed with... Multiply them to calculate the similarity. for The dimension value, i.e., the length of its vector, is here assumed to be 64-dimensional data processing; The feature vector at the current time; This represents the actual passenger flow demand at a historical point in time.

[0055] In step S3, real-time weather, time, and passenger data are input into the data extrapolation model. An intelligent optimization algorithm, combined with real-time traffic conditions, dynamically adjusts the allocation of public transportation resources, including real-time data analysis and intelligent adjustment, and resource allocation recommendations (e.g., on July 9th, a sunny evening rush hour. Based on historical data analysis, passenger flow is high during the evening rush hour on sunny days during the summer travel peak. Based on past data, it is estimated that 230 taxis and ride-hailing vehicles, as well as 12 airport buses, are needed at this time, along with the subway, to meet the rapid dispersal of passengers. However, the number of taxis, ride-hailing vehicles, and airport buses is insufficient at this time. By calculating the difference, the required number of taxis, ride-hailing vehicles, and airport buses can be output).

[0056] Real-time data analysis and intelligent adjustment include inputting key data such as real-time weather, time, and passenger numbers into the data extrapolation model to dynamically adjust and analyze the current scenario and predict the most likely allocation of public transportation resources in the future scenario.

[0057] Resource allocation recommendations include outputting the prediction results and pushing them to downstream third-party systems.

[0058] In step S4, the results of public transportation resource allocation are monitored in real time, and the results are continuously optimized and dynamically adjusted in combination with the feedback mechanism. The latest corrected results are pushed to the downstream third-party system, including monitoring data acquisition, allocation result correction, and data feedback.

[0059] Data acquisition includes real-time monitoring of airport public transportation, and timely identification of discrepancies between real-time data output and predicted data through manual intervention, adjusting the weight values ​​of each component accordingly. and error value ;

[0060] The correction of the allocation results includes: re-analyzing special data during the extrapolation and readjusting the assigned weights. and error value And correct the data results;

[0061] Data feedback includes outputting the latest correction results and pushing them to downstream third-party systems.

[0062] Another objective of this invention is to provide an airport public transportation pre-allocation processing system based on a big data model. This system implements the aforementioned airport public transportation pre-allocation processing method based on a big data model. The system includes:

[0063] Data acquisition module: used to acquire historical data from multiple sources, including flight information, passenger flow, weather conditions, transportation status, passenger arrival and departure information, and passenger travel intentions;

[0064] The data analysis and inference module, based on big data processing technology, uses machine learning and deep learning models to infer passenger demand and public transportation usage from historical data and generate data inference models.

[0065] The predictive scheduling module is used to input real-time weather, time and passenger data into the data extrapolation model, and uses intelligent optimization algorithms to dynamically adjust the allocation of public transportation resources in combination with real-time traffic conditions;

[0066] The intelligent monitoring and feedback module is used to monitor the allocation results of public transportation resources in real time, continuously optimize and dynamically adjust the results in conjunction with the feedback mechanism, and push the latest corrected results to downstream third-party systems.

[0067] Combining all the above technical solutions, the beneficial effects of this invention are as follows:

[0068] First, this invention utilizes real-time data analysis and intelligent optimization algorithms to achieve a more efficient traffic scheduling solution, which has significant practical implications and broad application prospects. The advantages of this invention also include: Accurate prediction: Based on multi-dimensional data analysis, it improves the accuracy of traffic demand forecasting. Dynamic scheduling: It adjusts traffic resources in real time, reducing passenger waiting time and improving traffic efficiency. Intelligent optimization: It adopts advanced algorithms LSTM and Transformer to improve resource utilization and reduce empty runs. Scalability: The system architecture supports access to multiple data sources and can adapt to the needs of different airports.

[0069] Secondly, this invention can reduce the pressure on public transportation near airports, improve passenger travel efficiency, and reduce passenger congestion at airports, thereby mitigating safety hazards caused by overcrowding. This invention can optimize public transportation scheduling, improve public transportation utilization, and thus reduce waste. It solves most of the management problems associated with arriving passengers at airports, significantly reducing secondary problems caused by passenger congestion at airports. Attached Figure Description

[0070] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure;

[0071] Figure 1 This is a flowchart of the airport public transportation pre-allocation processing method based on a big data model provided in an embodiment of the present invention;

[0072] Figure 2 This is the final curve graph generated from the historical data analysis provided in the embodiments of the present invention;

[0073] Figure 3 This is a schematic diagram of the airport public transportation pre-allocation processing method based on a big data model provided in an embodiment of the present invention;

[0074] Figure 4 This is a schematic diagram of an airport public transportation pre-allocation processing system based on a big data model, provided in an embodiment of the present invention. Detailed Implementation

[0075] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of the present invention. However, the present invention can be practiced in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0076] The innovation of this invention lies in the fact that it analyzes historical data to form a big data projection model that can handle airport public transportation, and uses real-time data analysis and intelligent optimization algorithms to generate more efficient traffic scheduling solutions.

[0077] Example 1: The airport public transportation pre-allocation processing method based on big data model provided in this embodiment of the invention has the following technical features.

[0078] Data collection and real-time analysis are achieved by connecting to the airport's production and operation system, passenger service system, airport environmental management system, and local municipal traffic management platforms. This allows for the acquisition of data such as flight information, passenger flow, passenger arrival and departure information, passenger travel intentions, weather conditions in the city where the airport is located, and traffic conditions in the city where the airport is located. A big data combined model using LSTM (Long Short-Term Memory) and Transformer (an attention-based model) is employed to analyze integrated historical and real-time data from multiple dimensions, including flight dynamics, passenger flow trends, and transportation status. The resulting real-time analysis results enable accurate prediction of airport public transportation demand.

[0079] Intelligent scheduling optimization: Utilizing machine learning algorithms and optimization models based on LSTM (Long Short-Term Memory Network) and Transformer (attention-based model), historical data is analyzed to generate algorithm models based on multi-dimensional airport data. Furthermore, traffic resource allocation can be dynamically adjusted based on real-time data, thereby improving the scientific nature and flexibility of scheduling schemes.

[0080] Enhancing the passenger travel experience: By using an algorithmic model generated based on multi-dimensional airport data, real-time data is analyzed to output predicted arrangements for airport public transportation, scheduling and adjusting the quantity of various public transportation options, thereby reducing passenger waiting time for public transportation at the airport, avoiding inconvenience caused by uneven resource allocation, and improving travel convenience and satisfaction.

[0081] Improve resource utilization: Based on big data analysis, the system can predict and arrange airport public transportation, which can rationally allocate taxis, buses, subways and other transportation tools, reduce idleness and overload, and improve overall operational efficiency.

[0082] This invention introduces intelligent and data-driven methods to provide airport public transportation systems with more accurate, efficient, and flexible allocation schemes, meeting the dual needs of efficient airport operation and comfortable travel for passengers.

[0083] Specifically, such as Figure 1 As shown in the figure, the airport public transportation pre-allocation processing method based on a big data model provided in this embodiment of the invention includes:

[0084] S1 acquires historical data from multiple sources, including flight information, passenger flow, weather conditions, transportation status, passenger arrival and departure information, and passenger travel intentions.

[0085] It includes data acquisition, data preprocessing, data clustering, and data distribution.

[0086] For example, data collection includes obtaining data such as flight information, passenger flow, passenger arrival and departure information, passenger travel intentions, weather conditions in the city where the airport is located, and traffic conditions in the city where the airport is located by connecting with platforms such as production and operation, passenger service, airport environmental management, and local traffic management.

[0087] Data preprocessing includes performing preliminary preprocessing on the collected data to filter out useless and erroneous data.

[0088] Data clustering involves grouping data based on correlation information such as the time period of occurrence and the passengers involved.

[0089] Data distribution involves distributing the processed and clustered data to downstream data analysis modules for further processing.

[0090] Another example is the need to collect historical public transportation data for the airport before establishing an accurate airport public transportation pre-allocation system. This data includes, but is not limited to, flight volume, passenger volume, weather conditions, airport traffic pressure, and airport public transportation (buses, subways, ride-hailing services, taxis) usage at various time points. This data can be obtained from airport operations platforms, passenger service platforms, airport integrated service management platforms, airport city traffic management platforms, and other platforms. By connecting data from various systems, the data is aggregated into a unified data pool, awaiting further processing. Figure 4 As shown;

[0091] S2, based on big data processing technology, uses machine learning and deep learning models to infer passenger demand and public transportation usage from historical data and generate data projection models.

[0092] This includes in-depth data processing, historical data analysis, and model deduction and correction.

[0093] For example, deep data processing includes: adjusting all historical data into a machine-readable format, such as weather conditions like sunny or rainy, and converting it into binary data that a computer can read. Another example is integrating data from the same time period, such as integrating weather, passenger, and flight data from 17:52 on January 7th, putting these data together, and converting them into binary data that a computer can read.

[0094] For example, historical data analysis includes: using models such as machine learning and deep learning to analyze all historical data to predict airport traffic usage at different times, in different weather conditions, and with different passenger numbers.

[0095] Setting time-related factors , The influencing factors change over time. Analysis of historical data shows a certain ratio between time and passenger volume: low passenger flow in the early morning, a larger morning peak, relatively stable daytime traffic, an evening peak, and then a decrease at night, ultimately forming a curve as shown below. Figure 2 As shown. Its formula is as follows:

[0096] (1)

[0097] In the formula, As a time-related factor, It is half the difference between the highest and lowest number of people on that day. Indicates average foot traffic. The current time;

[0098] Different weather conditions have a significant impact on passenger numbers and traffic. Weather can be mainly categorized into three types: weather type, weather pattern, ... One is the humidity, such as sunny or rainy weather. The last one is temperature. Based on historical data, the following formula was derived:

[0099] (2)

[0100] In the formula, For the usage of transportation (such as taxis), The weights of each influencing factor (e.g., for a certain 4F-level airport in China, the weight of its influence is...) It is 0.42. It is 0.12. It is 0.09. It is 0.11. It is 0.43. (0.29) Weather type For humidity, For temperature, This is the error value. For passenger flow, Indicate whether it is a holiday;

[0101] For example, model extrapolation and correction include: analyzing the predicted usage of public transportation in historical data and comparing it with the actual situation at the time, that is, substituting the actual weather, time and other data into formula (2), judging whether the result Y is similar to the extrapolation result. If they are similar, no adjustment is made; if they are not similar, adjustments are made. The value is adjusted to make the deduced result close to the final actual result. Repeating the above steps, if the deduced result consistently matches the actual result, the deduction model can be considered successful, and a data deduction model is generated. This model is implemented by writing code to execute the above formulas. It includes the implementation logic of the above formulas and the historical data set used for deduction.

[0102] Another example, such as Figure 3 , Figure 4 As shown, the generation of the data extrapolation model (data processing and model building) further includes:

[0103] (1) Perform preliminary processing on the collected data. First, use a computer to remove duplicate and null data. For data obtained from Excel, a Python script can be written, import the pandas library, and use the pd.read_excel method to read all the data in Excel. Iterate through each data item. If the key field is 'None', 'NULL', or an empty string, delete the current data. Finally, save all the data that has not been deleted. For data obtained from the database, a corresponding SQL script can be written to determine if the key field 'columns' is not null or an empty string. That is, use the encoding method to determine 'columns is not null or columns !=', and finally obtain all the data.

[0104] Then, through manual screening, the initially clear data is checked, and any abnormal data is processed and corrected to ensure data quality.

[0105] Feature extraction is performed on the processed data. First, useless feature values, such as "flight number" and "passenger name," are removed. Then, key features are marked and extracted. For example, "flight arrival time" can be used to determine if it is currently peak time, thus extracting the coverage area of ​​peak hours; "outdoor weather and precipitation" can be used to extract the impact of the weather on transportation; and "passenger's intention to come here" can be used to extract the likelihood of the passenger taking what kind of transportation later.

[0106] Dataset Construction: The processed data is used to construct a mapping set between these parameters and public transportation usage, such as: public transportation usage during peak flight times; public transportation usage during inclement weather (rain, strong winds, etc.); public transportation usage when passengers include tour groups, etc. These parameters also need to be combined into combined influencing parameters, such as: public transportation usage during off-peak hours and rainy days; public transportation usage when there are major events in the city / district and the weather is good; public transportation usage during off-peak hours and when there is no subway service, etc. Based on the construction, time features (flight arrival time, flight delay time, public transportation suspension time, etc.), flight information (number of flights, arrival / departure ratio, delay status, etc.), weather features (temperature, precipitation, wind speed, etc.), and historical traffic flow (passenger flow in the past N hours, taxi / subway utilization rate) are labeled as the main data feature set.

[0107] Training Model: For this invention, a big data combined model using LSTM (Long Short-Term Memory Network) and Transformer (attention-based model) is used as the data model (data inference model) of this system to handle short-term time dependence and long-term trends respectively, so as to obtain more accurate prediction results.

[0108] LSTM is primarily used to process short-term time series patterns, such as how flight changes over the past hour affect passenger flow. LSTM includes:

[0109] Forget Gate This determines how much past information to forget, such as forgetting nighttime data to focus on peak daytime demand. The formula is:

[0110] ;

[0111] In the formula, The activation function in LSTM. The input data includes current hourly flight information and weather conditions; The weights input at the previous time step. This refers to the data entered at the previous moment, including flight information and weather conditions. This is an offset value; the final calculation result is obtained by manually intervening in the numerical value. For the current moment, The previous moment; This is the weight of the current input. The importance of the input is judged manually and adjusted manually. The value is between 0 and 1. For example, if there are no flights at night, the required weight is smaller, so it is set to 0.1. This is the offset; the value is manually adjusted during the final calculation.

[0112] Input Gate This determines how the currently input information should be stored in the cell state, such as remembering the impact of severe weather on traffic. The formula is:

[0113] ;

[0114] In the formula, This is the weight of the current input gate. Here, the importance of the input is judged manually based on the input content and adjusted manually. The value is between 0 and 1. This is the input gate offset; the other parameters have the same meaning as the forget gate.

[0115] Cell State (This is the function of storing information in long-term memory, such as remembering past peak patterns. Its formula is:)

[0116] ;

[0117] The parameters have the same meaning as the input gate and forget gate.

[0118] Output gate This determines how much information about the current cell state is output, such as only outputting information that affects future passenger flow. The formula is:

[0119] ;

[0120] Final result This is the final result, representing the public transportation prediction for the next moment.

[0121] ;

[0122] This Long Short-Term Memory (LSTM) network model is coded in Python, using the torch.nn library to implement LSTM. The code is as follows:

[0123] import torch.nn as nn

[0124] class LSTMBlock(nn.Module):

[0125] def __init__(self, input_size, hidden_size, num_layers):super(LSTMBlock, self).__init__() self.lstm = nn.LSTM(input_size, hidden_size,num_layers, batch_first=1)

[0126] def forward(self, x): output, _ = self.lstm(x) return output

[0127] The input parameter input_size is the data for the input gate, hidden_size is the data for the forget gate, and num_layers is the data for each time period.

[0128] The Transformer is primarily used to capture long-term trends and global patterns, while LSTM is combined to learn short-term passenger flow changes, achieving accurate predictions. The formula is:

[0129] ;

[0130] In the formula, This is the result of the calculation formula used in this experiment; This is a normalization function that transforms numerical values ​​into a probability distribution. For matrix transpose, It is a matrix, and its transpose is the same as... Multiply to calculate similarity. for The dimension value, i.e. the length of its vector, is 64-dimensional data processing by default. Q (Query) represents the feature vector at the current time (such as the number of flights at 12:00 and the weather conditions), K (Key) represents a matrix (such as flights and passenger flow at 10:00 and 11:00), and V (Value) represents the actual passenger flow demand at historical time points.

[0131] The attention-based model is coded in Python using the torch.nn library, and its code is as follows:

[0132] class TransformerEncoder(nn.Module):

[0133] def __init__(self, d_model, nhead, num_layers): super(TransformerEncoder, self).__init__() self.encoder_layer =nn.TransformerEncoderLayer(d_model=d_model, nhead=nhead) self.transformer_encoder = nn.TransformerEncoder(self.encoder_layer, num_layers=num_layers)

[0134] def forward(self, x): return self.transformer_encoder(x)

[0135] Training and Optimization: Input all constructed datasets into the LSTM and Transformer models to output corresponding results. Manually validate the output results. If the validation results are unclear about actual public transportation usage, adjust the input parameters and check if the output data closely matches the actual results. Calculate the variance of the data. This refers to data predicted by LSTM and Transformer. This represents the actual data, and n represents the number of iterations. When the variance is less than a reasonable value, the training optimization is considered successful.

[0136] ;

[0137] S3 inputs real-time weather, time, and passenger data into the data extrapolation model, and uses intelligent optimization algorithms to dynamically adjust the allocation of public transportation resources in conjunction with real-time traffic conditions;

[0138] Mature intelligent optimization algorithms, such as Genetic Algorithm (GA) and Ant Colony Algorithm (ACO), can be used here to dynamically adjust the allocation of public transportation resources in conjunction with real-time traffic conditions. This includes real-time data analysis for intelligent adjustment and resource allocation recommendations.

[0139] For example, real-time data analysis and intelligent adjustment include inputting key data such as real-time weather, time, and passenger numbers into a data extrapolation model to dynamically adjust and analyze the current scenario and predict the most likely allocation of public transportation resources in future scenarios.

[0140] Resource allocation recommendations include outputting the prediction results and pushing them to downstream third-party systems.

[0141] Another example is that by inputting future data into a pre-trained data extrapolation model, prediction results can be output and then distributed to other third-party systems.

[0142] When the actual time arrives, if the deviation between the actual result and the predicted result is significant, record the environmental parameters of the current time period and analyze the reasons for the current deviation. Investigate for any interfering factors and adjust the parameter values ​​in both the LSTM (Long Short-Term Memory) and Transformer (attention-based model) models accordingly. Figure 3 , Figure 4 As shown.

[0143] S4 monitors the allocation of public transportation resources in real time, continuously optimizes and dynamically adjusts the results based on feedback mechanisms, and pushes the latest corrected results to downstream third-party systems.

[0144] This includes monitoring data acquisition, allocation result correction, and data feedback.

[0145] For example, monitoring data acquisition includes real-time monitoring of airport public transportation and other data. Inconsistencies between the real-time data output and the predicted data are identified and addressed manually, and the weights of each component are adjusted accordingly. and error value .

[0146] For example, the correction of the allocation results includes: if the impact of tornadoes was not considered in the deduction of steps S2-S3, but tornadoes or other extremely special conditions occurred in the actual results, then this special data will be re-analyzed and its weights will be readjusted. and error value And correct the data results.

[0147] For example, data feedback includes outputting the latest correction results and pushing them to downstream third-party systems.

[0148] Another example is the use of appropriate monitoring methods (such as manual spot checks and video analysis) to compare and contrast the predicted results of the data extrapolation model with the actual values ​​required for public transportation. This allows for the identification and recording of cases where data deviations are significant and inconsistent with the ideal results. Once a sufficient amount of data has been collected to form a dataset, it is analyzed step by step to identify any unrelated environmental factors that could influence the model, thereby optimizing the overall model.

[0149] Example 2, as Figure 4 The diagram shows the schematic of the airport public transportation pre-allocation processing system based on a big data model provided in this embodiment of the invention.

[0150] Specifically, the airport public transportation pre-allocation processing system based on a big data model provided in this embodiment of the invention includes:

[0151] The data acquisition module is used to obtain multi-source data such as flight information, passenger flow, weather conditions, transportation status, passenger arrival and departure information, and passenger travel intentions. It includes sub-modules for data acquisition, data preprocessing, data clustering, and data distribution. The data acquisition sub-module connects to platforms such as production operations, passenger services, airport environmental management, and local traffic management to obtain data on flight information, passenger flow, passenger arrival and departure information, passenger travel intentions, weather conditions in the city where the airport is located, and traffic conditions in the city where the airport is located. The data preprocessing sub-module performs preliminary preprocessing on the acquired data, filtering out useless and erroneous data. The data clustering sub-module groups the data according to correlation information such as time period and passenger affiliation. The data distribution sub-module distributes the processed and clustered data to downstream data analysis modules for further processing.

[0152] The data analysis and prediction module includes models based on big data processing technology, such as machine learning and deep learning, to predict passenger demand and public transportation usage from historical data. This includes sub-modules such as deep data processing, historical data analysis, and model deduction and correction. For example, the historical data analysis sub-module analyzes all historical data using machine learning and deep learning models to predict airport traffic usage under different times, weather conditions, and passenger volumes. The model deduction and correction sub-module compares the predicted public transportation usage from historical data with the actual situation at the time. It substitutes actual weather and time data into the formula and determines whether the result Y is close to the predicted result. If close, no adjustment is made; if not, the weight values ​​are adjusted to make the predicted result closer to the final actual result. This process is repeated until the predicted result consistently matches the actual result, at which point the deduction model is considered successful and a data deduction model is generated.

[0153] The predictive scheduling module uses intelligent optimization algorithms combined with real-time traffic conditions to dynamically adjust the allocation of public transportation resources. It includes sub-modules such as real-time data analysis for intelligent adjustment and resource allocation recommendation. For example, the real-time data analysis for intelligent adjustment sub-module inputs key data such as real-time weather, time, and passenger numbers into the projection model to analyze the current scenario and predict the most likely public transportation usage in the future. The resource allocation recommendation sub-module outputs the prediction results and pushes them to downstream third-party systems.

[0154] The intelligent monitoring and feedback module is used to monitor traffic allocation effectiveness in real time and continuously optimize scheduling strategies through a feedback mechanism. It includes sub-modules for monitoring data acquisition, allocation result correction, and data feedback.

[0155] For example, the monitoring data acquisition submodule monitors airport public transportation and other data in real time. It manually identifies discrepancies between the real-time data output and the predicted data, adjusting the weights and error values ​​of each component. The result correction submodule, for instance, if the prediction did not consider the impact of tornadoes but actual conditions such as tornadoes occurred, reanalyzes this special data, readjusts its weights and error values, and corrects the data results. The data feedback submodule outputs the latest predicted results and pushes them to downstream third-party systems.

[0156] To further illustrate the effects of the embodiments of the present invention, the following experiments were conducted.

[0157] Application Example 1: Predicting airport bus scheduling based on flight data and passenger flow.

[0158] The required input parameters include:

[0159] Flight information: arrival time, flight number, delay status, etc.

[0160] Passenger flow information: Real-time passenger numbers and travel preferences (such as taxis, buses, etc.) within the airport.

[0161] Weather conditions: Heavy rain, heavy snow and other severe weather may affect traffic demand.

[0162] Traffic conditions: Current availability of taxis and buses around the airport, and traffic congestion.

[0163] This invention can obtain information from passenger service and operational platforms regarding the number of flights and passengers in the near future. It then considers outdoor weather conditions and historical data for similar weather during this period to estimate the number of airport buses needed to serve these passengers. Furthermore, it utilizes IoT devices to obtain data on the number of passengers currently stranded in the terminal and traffic congestion around the airport. Finally, this data is fed into LSTM and Transformer models to output the estimated number of buses required to meet passenger demand under these conditions.

[0164] Application Example 2: Based on long-term data, predict the corresponding bus adjustments when airport flight schedules change seasonally.

[0165] The required input parameters include: Flight Information: Upcoming flight schedules, historical flight adjustments for this quarter, etc. Passenger Information: Passenger information from previous years. Weather Conditions: Historical weather conditions for this quarter. Holidays and Major Events: Information on holidays and major events included in this quarter.

[0166] This invention uses historical passenger information, flight schedules, and flight counts to predict daily passenger numbers for the next quarter. It then considers weather conditions and the presence of holidays or major events to determine if there are other influencing factors. Finally, the data is fed into LSTM and Transformer models to output the daily demand for airport buses, ride-hailing services, etc., for the entire next quarter.

[0167] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by any person skilled in the art within the scope of the technology disclosed in the present invention, and within the spirit and principles of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for pre-allocation of airport public transportation based on a big data model, characterized in that, The method includes the following steps: S1 acquires historical data from multiple sources, including flight information, passenger flow, weather conditions, transportation status, passenger arrival and departure information, and passenger travel intentions. S2, based on big data processing technology, uses machine learning and deep learning models to infer passenger demand and public transportation usage from historical data and generate data projection models. S3 inputs real-time weather, time, and passenger data into the data extrapolation model, and uses intelligent optimization algorithms to dynamically adjust the allocation of public transportation resources in conjunction with real-time traffic conditions; S4 monitors the allocation results of public transportation resources in real time, continuously optimizes and dynamically adjusts the results based on the feedback mechanism, and pushes the latest corrected results to downstream third-party systems. In step S2, based on big data processing technology, historical data is collected, cleaned, and analyzed to remove abnormal interference data, resulting in a set of historical data suitable for machine learning and deep learning. Then, machine learning and deep learning models are used to infer passenger demand and public transportation usage from the historical data, generating a data inference model; including: historical data analysis, model inference and correction. Historical data analysis includes: using machine learning and deep learning models to analyze all historical data and predict airport traffic usage under different times, weather conditions, and passenger numbers. Set the time-influencing factor to the current time. , The influencing factors change over time, as expressed by: （1） In the formula, As a time-related factor, It is half the difference between the highest and lowest number of people on that day. Indicates average foot traffic; Based on historical data, the formula evolved as follows: （2） In the formula, For vehicle usage, The weights of each influencing factor, Weather type For humidity, For temperature, This is the error value. For passenger flow, Indicate whether it is a holiday; Model extrapolation and correction include: analyzing the predicted usage of public transportation in historical data, comparing it with the actual situation at the time, substituting real-time weather, time, and passenger data into formula (2), and judging the results. If the results are similar to the predicted results, no adjustments are made; otherwise, adjustments are made. The value is adjusted to make the deduction result close to the final actual result; Repeat the above steps. When the deduction result is consistently close to the actual result, the deduction model is considered successful, and a data deduction model is generated. This data deduction model is implemented by writing code to realize the above formula and includes the implementation logic of the above formula as well as a set of historical data for deduction. In step S4, the results of public transportation resource allocation are monitored in real time, and the results are continuously optimized and dynamically adjusted in combination with the feedback mechanism. The latest corrected results are pushed to the downstream third-party system, including monitoring data acquisition, allocation result correction, and data feedback. Data acquisition includes real-time monitoring of airport public transportation, and timely identification of discrepancies between real-time data output and predicted data through manual intervention, adjusting the weight values ​​of each component accordingly. and error value ; The correction of the allocation results includes: re-analyzing special data during the extrapolation and readjusting the assigned weights. and error value And correct the data results; Data feedback includes outputting the latest correction results and pushing them to downstream third-party systems.

2. The airport public transportation pre-allocation processing method based on a big data model according to claim 1, characterized in that, In step S1, historical data from multiple sources, including flight information, passenger flow, weather conditions, transportation status, passenger arrival and departure information, and passenger travel intentions, are acquired, including: data collection, data preprocessing, data clustering, and data distribution. Data collection involves connecting with production and operations, passenger services, airport environmental management, and local traffic management platforms to obtain data on flight information, passenger flow, passenger arrival and departure information, passenger travel intentions, weather conditions in the city where the airport is located, and traffic conditions in the city where the airport is located. Data preprocessing involves performing preliminary processing on the collected data to filter out useless and erroneous data. Data clustering is the process of grouping data according to the time period and the related information of the passengers involved. Data distribution involves distributing processed and clustered data to downstream processes.

3. The airport public transportation pre-allocation processing method based on a big data model according to claim 1, characterized in that, In step S3, real-time weather, time, and passenger data are input into the data extrapolation model. Intelligent optimization algorithms are used to dynamically adjust the allocation of public transportation resources in combination with real-time traffic conditions, including real-time data analysis, intelligent adjustment, and resource allocation recommendation. Real-time data analysis and intelligent adjustment involves inputting key data such as real-time weather, time, and passenger numbers into a data extrapolation model to dynamically adjust and analyze the current scenario and predict the allocation of public transportation resources in future scenarios. Resource allocation recommendations include outputting the prediction results and pushing them to downstream third-party systems.

4. An airport public transportation pre-allocation processing system based on a big data model, characterized in that, The system implements the airport public transportation pre-allocation processing method based on a big data model as described in any one of claims 1-3, and the system includes: Data acquisition module: used to acquire historical data from multiple sources, including flight information, passenger flow, weather conditions, transportation status, passenger arrival and departure information, and passenger travel intentions; The data analysis and inference module, based on big data processing technology, uses machine learning and deep learning models to infer passenger demand and public transportation usage from historical data and generate data inference models. The predictive scheduling module is used to input real-time weather, time and passenger data into the data extrapolation model, and uses intelligent optimization algorithms to dynamically adjust the allocation of public transportation resources in combination with real-time traffic conditions; The intelligent monitoring and feedback module is used to monitor the allocation results of public transportation resources in real time, continuously optimize and dynamically adjust the results in conjunction with the feedback mechanism, and push the latest corrected results to downstream third-party systems.

Images