CN118467932B - Civil aviation important activity resisting information processing method and system - Google Patents

Abstract

The invention belongs to the technical field of civil aviation information processing, and discloses a method and a system for processing important civil aviation activity resisting information. The method comprises the steps of collecting and preprocessing a data source, and extracting and modeling characteristics of data of the data preprocessing data source; training and optimizing a prediction model; performing result verification and correction on the processed result; and continuously training and optimizing the prediction model, and adapting to continuously changing data and requirements. The method can better capture the time sequence characteristics and rules in the important activities of civil aviation against information data, and improves the accuracy and stability of prediction. The method can provide important data analysis and prediction tools for civil aviation companies, airport management departments and the like, help the civil aviation companies to better schedule flights and manage resources, and improve the safety and efficiency of civil aviation transportation.

Description

A method and system for processing arrival and departure information of important civil aviation activities

Technical Field

The present invention belongs to the technical field of civil aviation information processing, and in particular relates to a method and system for processing arrival and departure information of important civil aviation activities.

Background Art

With the development of the global economy and the improvement of people's living standards, the civil aviation industry has experienced rapid growth in the past few decades. Civil aviation transportation has become one of the main ways for people to cross borders and quickly reach their destinations. A large number of flights take off and land around the world every day, providing convenience for business travel, tourism and cargo transportation.

In civil aviation transportation, accurate flight arrival and departure information is crucial for airlines, airport management departments and passengers. The arrival and departure times of flights determine the normal operation of flights, which directly affects the itinerary arrangements of passengers and the logistics scheduling of goods. Any errors or delays in arrival and departure information may cause unnecessary troubles and losses, so ensuring the accuracy and timeliness of arrival and departure information is crucial to the operation of the entire civil aviation system.

Traditional methods of processing arrival and departure information mainly rely on manual or rule-based automated systems. These methods often rely on manual data entry and processing by personnel, which has the following problems:

Inaccurate information: Due to human factors or uncertainty in data sources, traditional methods often cannot guarantee the accuracy of arrival and departure information.

Inefficient processing: Manual processing requires a lot of time and manpower and cannot meet the needs of real-time processing, especially during peak hours or when emergencies occur.

Difficult to cope with complex situations: The amount of data involved in the civil aviation system is huge and complex, and traditional methods are often unable to effectively handle large-scale data and complex situations.

With the development of information technology, artificial intelligence technologies such as big data and deep learning have been widely used in various fields, providing new ideas and solutions to solve the problems of traditional methods. In the field of civil aviation, big data technology can help collect and process massive amounts of flight data, airport data and weather data, providing more comprehensive and accurate information support. Deep learning technology can automatically extract features from massive amounts of data and make predictions and decisions through training models, providing new possibilities for real-time processing of arrival and departure information.

In view of the rapid development of the civil aviation industry and the limitations of traditional methods, a more advanced and efficient arrival and departure information processing method is urgently needed to meet the growing demand.

Summary of the invention

In order to overcome the problems existing in the related technologies, the disclosed embodiments of the present invention provide a method and system for processing arrival and departure information of important civil aviation activities, and specifically relate to a method for processing arrival and departure information of important civil aviation activities. The present invention responds to actual needs based on big data and deep learning. It can make full use of existing data resources and combine with advanced artificial intelligence technology to improve the accuracy, real-time and efficiency of arrival and departure information processing, injecting new vitality into the development of the civil aviation industry.

The technical solution is as follows: A method for processing arrival and departure information of important civil aviation activities, comprising:

S1, collecting and preprocessing data sources; the data sources include: flight information, airport information, weather information, personnel arrival and departure information, personnel destination information, and nearby highway traffic information;

S2, extracting features and modeling the data from the data source; feature extraction includes: extracting indicators or attributes that reflect data features and information from the original data, and inputting the feature-extracted data into the arrival and departure information prediction model established using a deep learning algorithm;

S3: Train and optimize the arrival and departure information prediction model. After the training and optimization is completed, the arrival and departure information is predicted and processed, and the processing results are output to the platform for query and use;

S4, verifying and correcting the processed results. When the arrival and departure information prediction model predicts deviations or errors, correct the results through manual intervention or adjustment of the arrival and departure information prediction model parameters;

S5, continuously train and optimize the arrival and departure information prediction model, and continuously adjust the weight ratio of each input parameter to adapt to the ever-changing data and needs;

S6, puts the trained arrival and departure information prediction model into practice to continuously provide decision-making suggestions to decision makers.

In step S1, the data source is collected, including: manually uploading, extracting the interface provided by the third-party business system, directly obtaining from the third-party business system database, and crawling data; and the collected data is saved in the database;

Preprocessing of data sources includes:

(1) Missing value processing: Identify and process missing values in the data and fill in the missing data; for data with similar references, make similar comparisons and fill in the missing values; for data without similar references, delete the current data to avoid the current data affecting the subsequent training results;

(2) Outlier detection: Delete the data that is obviously abnormal; adjust the data if there are abnormal conditions;

(3) Data format conversion: The data format is adjusted uniformly to facilitate subsequent data import. The date and time format is unified into a standard timestamp format, all text is changed into character format, and the numerical values retain the corresponding number of digits;

(4) Data deduplication: Classify and organize data, compare similar data, identify and delete completely identical data, and ensure data uniqueness and consistency;

(5) Data standardization: Organize the data to be standardized and establish a standardized reference database, convert the data into the corresponding data in the standardized database, and comply with the unified data standards and specifications, including but not limited to: adjusting the airline logos of all flights to the corresponding three-letter codes; adjusting all data time to Beijing time;

In step S1, after the data source is preprocessed, the quality of the processed data needs to be evaluated.

In step S2, feature selection is performed before feature extraction. The field of feature selection is defined based on the frequency of occurrence of the field and the artificially defined importance. The expression is:

;

In the formula, Determine the value for the eigenvalue, is the frequency of occurrence, To define the importance of Values are between 0-1, very important ,unimportant ; is the total amount of data;

When the calculation results When it is greater than 0.6, the current data is considered to be characteristic, and the current value is selected as the characteristic value; the most representative and distinguishing feature subset is selected from the original data, and the feature subset includes but is not limited to:

Flight information features: including flight departure time, arrival time, flight number, and aircraft model information;

Airport information features: including airport code, terminal, and boarding gate information;

Weather information characteristics: including temperature, humidity, wind direction and speed meteorological factors;

Other information features: including the destination the passenger is travelling to, how the passenger is travelling to the destination, and subsequent plans.

In step S2, the arrival and departure information prediction model established includes:

Dataset construction: Divide the previously collected, cleaned and feature-extracted data into three parts: one for training, one for validation, and one for testing.

Build a long short-term memory (LSTM) model: Use the LSTM model as the underlying layer of deep learning, select multiple LSTM layers to increase the depth of the model, and stack the LSTM layers together in sequence to form a deep LSTM network.

In step S3, the arrival and departure information prediction model is trained and optimized, including: the LSTM training process, using the back propagation algorithm and the stochastic gradient descent optimization algorithm; during the training process, the gradient of the loss function with respect to the weight is calculated, and the weight is updated using the stochastic gradient descent optimization algorithm to minimize the loss function;

After the prediction model is trained and optimized, the arrival and departure information prediction and processing include: after the arrival and departure information prediction model is trained, evaluation and tuning are performed. During the evaluation process, the data in the test set is used to evaluate the prediction performance of the model on unknown data; the evaluation indicators include mean square error (MSE), mean absolute error (MAE), and accuracy; they are used to measure the prediction accuracy and generalization ability of the arrival and departure information prediction model; if the three types of data are within the applicable range, the arrival and departure information prediction model passes; if the performance of the arrival and departure information prediction model is inaccurate, it is optimized by adjusting the model structure and adjusting the hyperparameters.

The mean square error (MSE) is used to measure the degree of match between the predicted value and some true value. The expression is:

;

In the formula, is the number of samples, For real data, The predicted data represents the predicted data.

Furthermore, the mean absolute error (MAE) is used to calculate the sample standard deviation of the difference between the predicted and observed values, expressed as:

;

In the formula, is the number of samples.

In step S4, the processing result is verified and corrected, including: first manually calculating the result that the training data should have output, and then comparing the output result of the arrival and departure information prediction model with the manually calculated result.

Another object of the present invention is to provide a system for processing arrival and departure information of important civil aviation activities, and to implement the method for processing arrival and departure information of important civil aviation activities, the system comprising:

Data collection and preprocessing module: collect and preprocess data sources;

Feature extraction module: extracts representative and discriminative features from preprocessed data for deep learning model training and prediction;

Model training module: Use deep learning algorithms to train the extracted features and build an arrival and departure information prediction model;

Information processing and prediction module: Apply the trained LSTM model to the new arrival and departure information data of important civil aviation activities for information processing and prediction.

The feature extraction module is also used to select the most representative and discriminative feature subsets from the original data, including flight information features, airport information features, weather information features, and other information features;

The model training module is used to construct the arrival and departure information prediction model, including data set construction, construction of long short-term memory (LSTM) model, model training and optimization, and model evaluation and tuning.

Combining all the above technical solutions, the beneficial effects of the present invention are as follows: the present invention can comprehensively optimize the collection and analysis process of arrival and departure information to ensure the safe and orderly operation of important activities, and can also improve the operational efficiency of airports and other places. The system stores and analyzes information on important activities and arrival and departure personnel, combines big data, the Internet of Things and other technical means, and through the collaboration of multiple systems, provides timely and accurate decision support for airports and event organizers. It is committed to promoting the digital and intelligent development of civil aviation and improving the safety and service levels of the entire industry.

Compared with the prior art, the present invention improves the intelligence level of civil aviation transportation information processing, can accurately predict the arrival and departure of important civil aviation activities in the future, and provide important decision-making support for civil aviation transportation. It fully explores the potential laws and characteristics in the arrival and departure information data of important civil aviation activities, realizes the precise processing and prediction of civil aviation transportation information, and improves the efficiency and accuracy of information processing. It adopts a combination of big data and deep learning, which can better capture the time series characteristics and laws in the arrival and departure information data of important civil aviation activities, and improves the accuracy and stability of the prediction. It can provide important data analysis and prediction tools for civil aviation companies, airport management departments, etc., help them better carry out flight scheduling and resource management, and improve the safety and efficiency of civil aviation transportation.

The present invention can greatly optimize the management of passengers by airport and flight management personnel, and reduce the evacuation pressure of airport personnel through reasonable optimization schemes. Through reasonable evacuation schemes, the investment of labor costs and other resource costs can be reduced; management suggestions for event participants can be provided to event organizers, thereby reducing event costs; early warnings can be provided to local transportation departments in advance, so that future traffic forecast pressures can be grasped in a timely manner, and traffic management plans can be formulated in a timely manner; appropriate travel plans can be provided to event participants to avoid problems such as traffic rush hours that affect travel.

BRIEF DESCRIPTION OF THE DRAWINGS

The accompanying drawings herein are incorporated in and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description, serve to explain the principles of the present disclosure;

1 is a flow chart of a method for processing arrival and departure information of important civil aviation activities provided by an embodiment of the present invention;

2 is a schematic diagram of a method for processing arrival and departure information of important civil aviation activities provided by an embodiment of the present invention;

3 is a schematic diagram of a system for processing arrival and departure information of important civil aviation activities provided by an embodiment of the present invention;

FIG4 is a diagram of an LSTM model stacked using multiple LSTM layers provided in an embodiment of the present invention;

FIG5 is a processing flow chart of the overall LSTM model provided by an embodiment of the present invention;

In the figure: 1. Data acquisition and preprocessing module; 2. Feature extraction module; 3. Model training module; 4. Information processing and prediction module.

DETAILED DESCRIPTION

In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention, so the present invention is not limited by the specific implementation disclosed below.

The innovation of this invention is that it uses deep learning to process and analyze the massive amount of civil aviation arrival and departure information, and provides passenger management suggestions to event managers, airports and other relevant departments, thereby improving the operational efficiency of airports and transportation departments.

Embodiment 1, as shown in FIG1 , the method for processing arrival and departure information of important civil aviation activities provided by the embodiment of the present invention includes:

S1, collecting and preprocessing data sources; the data sources include: flight information, airport information, weather information, personnel arrival and departure information, personnel destination information, and nearby highway traffic information;

S2, extracting features and modeling the data from the data source; feature extraction includes: extracting indicators or attributes that reflect data features and information from the original data, and inputting the feature-extracted data into the arrival and departure information prediction model established using a deep learning algorithm;

S3: Train and optimize the arrival and departure information prediction model. After the training and optimization is completed, the arrival and departure information is predicted and processed, and the processing results are output to the platform for query and use;

S4, verifying and correcting the processed results. When the arrival and departure information prediction model predicts deviations or errors, correct the results through manual intervention or adjustment of the arrival and departure information prediction model parameters;

S5, continuously train and optimize the arrival and departure information prediction model, and continuously adjust the weight ratio of each input parameter to adapt to the ever-changing data and needs;

S6, puts the trained arrival and departure information prediction model into practice to continuously provide decision-making suggestions to decision makers.

Exemplarily, in step S1, data collection and preprocessing: In order to establish an accurate and reliable arrival and departure information processing system, large-scale data collection and preprocessing are first required. The data sources cover flight information, airport information, weather information, personnel arrival and departure information, personnel destination information, nearby road traffic information and other aspects. These data come from the systems of various airlines, the airport's operation management system, weather forecast agencies and personnel questionnaires, etc. In the collection process, the integrity, accuracy and timeliness of the data must be taken into account. After the data collection is completed, preprocessing operations are required, including data cleaning, deduplication, outlier processing, etc. The purpose of this step is to ensure the quality of the data, eliminate noise and anomalies in the data, and prepare for subsequent feature extraction and model training.

In step S2, feature selection is performed before feature extraction. The field of feature selection is defined based on the frequency of occurrence of the field and the artificially defined importance. The expression is:

;

In the formula, Determine the value for the eigenvalue, is the frequency of occurrence, To define the importance of Values are between 0-1, very important ,unimportant ; is the total amount of data;

When the calculation results When it is greater than 0.6, the current data is considered to be characteristic, and the current value is selected as the characteristic value; the most representative and distinguishing feature subset is selected from the original data, and the feature subset includes but is not limited to:

Flight information features: including flight departure time, arrival time, flight number, and aircraft model information;

Airport information features: including airport code, terminal, and boarding gate information;

Weather information characteristics: including temperature, humidity, wind direction and speed meteorological factors;

Other information features: including the destination the passenger is travelling to, how the passenger is travelling to the destination, and subsequent plans.

In step S2, feature extraction and data modeling: after data preprocessing is completed, the data needs to be feature extracted and modeled. Feature extraction refers to extracting indicators or attributes that can reflect data characteristics and information from the original data to provide input for model training. In the processing of flight arrival and departure information, multiple features may be involved, such as the flight's departure time, arrival time, flight number, airport code, weather conditions, destination, and mode of transportation after arrival. After extracting the features, the next step is to establish a prediction model. In the present invention, deep learning algorithms are used for modeling, including convolutional neural networks (CNN), recurrent neural networks (RNN), etc. Through training on a large amount of historical data, the model can learn the potential laws and characteristics in the data, thereby realizing the prediction and analysis of arrival and departure information.

It is understandable that the LSTM model (Long Short-Term Memory Network Model) is chosen here. It is a variant of the recurrent neural network (RNN) that aims to solve the gradient vanishing and gradient exploding problems of traditional RNN when processing long sequences. LSTM introduces a special storage unit and gating mechanism to more effectively capture and process long-term dependencies in sequence data.

In step S3, real-time information processing and result output: After the model training is completed, it can be applied to the actual arrival and departure information processing process. The system will receive real-time data from data sources such as airlines and airports in real time, and predict and process arrival and departure information through pre-trained models. This process needs to take into account the real-time and accuracy of the data to ensure that the system can respond and handle various situations in a timely manner. After the processing is completed, the system will output the results to the relevant system or platform for airlines, airport management departments and passengers to query and use. These results may include information such as the estimated arrival time of the flight, the actual arrival time, delays, and recommended transportation methods to the destination, providing reference and decision-making basis for event organizers and other business departments.

Exemplarily, in step S3, training and optimizing the arrival and departure information prediction model includes: a training process of LSTM, using a back propagation algorithm and a stochastic gradient descent optimization algorithm; during the training process, calculating the gradient of the loss function with respect to the weight, and updating the weight using the stochastic gradient descent optimization algorithm to minimize the loss function;

Specifically, it includes: using the historical arrival and departure information data of important civil aviation activities to learn and optimize the parameters of the LSTM model used. During the training process, the prepared training set arrival and departure information data is input into the LSTM network, and then the model is trained and optimized using the validation set data. The model parameters are continuously adjusted through the back propagation algorithm and the LSTM optimizer to reduce the loss function. For example, 10,000 sets of data are prepared in advance, of which 9,500 data are used as training sets and 500 data are used as validation sets. First, the training set is input into the LSTM network and the result model is output. Then, for the 500 validation sets, the results that should be output are manually judged. The validation set is input into the model again, and the results of the validation set output are compared with the results of manual judgment. If the comparison results are not ideal, the model parameters are adjusted, and then simulation tests are performed until the similarity between the output results and the results of manual judgment reaches 95%.

In step S4, verifying and correcting the processing result includes: first manually calculating the result that the training data should have output, and then comparing the output result of the arrival and departure information prediction model with the manually calculated result;

When deviations or errors occur in the arrival and departure information prediction model, the results are corrected through manual intervention or adjustment of model parameters.

For example, if the arrival and departure information prediction model outputs the wrong estimated time of arrival at the airport, the weights of the weather and the airline to which the flight belongs on the time are adjusted so that the final output result is close to the result calculated manually. For another example, if the arrival and departure information prediction model has a large error in calculating the time a passenger will stay at the airport, the weights of the passenger information on the stay information are adjusted so that the final output result is close to the result calculated manually.

For example, result verification and correction: In order to ensure that the arrival and departure information output by the system is accurate and reliable, result verification and correction are also required. This process may involve comparing with the actual situation, verifying the prediction results, and correcting them according to the actual situation. For example, when the arrival and departure information prediction model predicts deviations or errors, the results can be corrected by manual intervention or adjustment of the arrival and departure information prediction model parameters to improve the accuracy and credibility of the system.

In step S5, continuous optimization and updating: With the development of the aviation transportation industry and the continuous accumulation of data, the arrival and departure information processing system also needs to be continuously optimized and updated. This includes continuous training and optimization of the model to adapt to the ever-changing data and needs. At the same time, it is also necessary to pay attention to the development and application of new technologies, and timely introduce and apply new algorithms and methods, such as the Adam optimization algorithm and gradient clipping technology, to improve the performance and efficiency of the system.

As shown in FIG. 2 , as another embodiment of the present invention, the method for processing arrival and departure information of important civil aviation activities provided in an embodiment of the present invention includes: data collection and preprocessing, feature extraction and data modeling, real-time information processing and result output, result verification and correction, and continuous optimization and updating.

Embodiment 2, as shown in FIG3, the important civil aviation activity arrival and departure information processing system provided by the embodiment of the present invention includes:

Data collection and preprocessing module 1: collect and preprocess data to ensure the accuracy and reliability of subsequent processing. The main steps of this module are as follows:

S101, data collection: the data to be collected can be manually uploaded, extracted from the interface provided by the third-party business system, directly obtained from the third-party business system database, data crawling, etc. As for the collection method, it is necessary to communicate and connect with the third-party system and use the method provided by the third-party system for collection, so the present invention will not elaborate on it.

S102, data storage: the collected data is saved in the database through the saving function of the database software. This method is a function provided by the database software, so it is not described in detail in the present invention.

S103, data cleaning and preprocessing: Because the collected data may be different and the accuracy of the data cannot be guaranteed, human intervention is required at this time. Review and adjust the data, including at least the following aspects:

(1) Missing value processing: Identify and process missing values in the data and fill in the missing data; for data with similar references (for example, the same airline, the same departure and destination stations, and the same flight model, which can be considered similar and can be used as a reference), perform similar comparisons and fill in the missing values; for data without similarity, delete the current data to prevent the current data from affecting the subsequent training results;

(2) Outlier detection: For data with obvious anomalies, delete them uniformly (obvious anomalies refer to data that will seriously interfere with training. For example, a flight normally needs to fly for 2 hours, but one day it makes an emergency landing or returns due to abnormal circumstances. Such data is not conducive to model training and needs to be excluded); for data with minor anomalies, adjust the data (for example, a flight normally flies for 2 hours, but on that day due to factors such as heavy fog, the plane flew for 2.5 hours. Such data can be manually adjusted to change the flight time from 2.5 hours to 2 hours);

(3) Data format conversion: Convert the collected data into a unified format for subsequent processing. For example, the date and time format can be unified into a standard timestamp format to facilitate time series analysis and processing.

(4) Data deduplication: Identify and delete duplicate data records to ensure data uniqueness and consistency.

(5) Data standardization: Standardize the data to make it conform to unified data standards and specifications to facilitate subsequent data analysis and modeling. For example, the airline logos of all flights are changed to their corresponding three-letter codes; all data times are adjusted to use standard time as the benchmark, etc.

S104, after data cleaning and preprocessing are completed, the processed data needs to be evaluated for quality, including evaluation of data integrity, accuracy, and consistency to ensure that the data quality meets the requirements.

Feature extraction module 2: The feature extraction module is one of the key links in the present invention. Its main task is to extract representative and discriminative features from the preprocessed data for deep learning model training and prediction. Before feature extraction, feature selection is required. Feature selection refers to selecting the most representative and discriminative feature subsets from the original data to reduce data dimensions, improve model efficiency and reduce the risk of overfitting. In the processing of flight arrival and departure information, features related to the flight status can be selected according to domain knowledge and actual needs, and these feature values can be extracted, for example:

Flight information features: including flight departure time, arrival time, flight number, aircraft type, etc. These features can reflect the basic attributes and status of the flight and have an important impact on the prediction of flight arrival and departure information.

Airport information features: including airport code, terminal, boarding gate, etc. These features can reflect the operation status of the airport and the operation of the flight, and have a certain impact on the accuracy of flight arrival and departure information.

Weather information features: including meteorological factors such as temperature, humidity, wind direction and speed. Weather has an important impact on flight operations, so incorporating weather information into the scope of feature extraction will help improve the accuracy and reliability of flight arrival and departure information.

Other information features: including the destination of the passenger, the method of going to the destination, and subsequent plans. This information helps to analyze and organize the subsequent plans of the personnel and provide corresponding suggestions for other departments.

Model training module 3: Model training is one of the core links in this invention. Its main task is to use deep learning algorithms to train the extracted features and build an arrival and departure information prediction model. In the processing of flight arrival and departure information, the goal of model training is to predict the arrival and departure information of future flights by learning the patterns and rules in historical data. This module mainly includes the following steps:

S301, data set construction: refers to the data collected, cleaned and feature extracted in the early stage. The data is divided into three parts, one is the training set, one is the validation set, and one is the test set.

S302, constructing an LSTM (Long Short-Term Memory) model: In the present invention, the LSTM model is used as the bottom layer of deep learning. Multiple LSTM layers are selected to increase the depth of the model, and these LSTM layers are stacked together in sequence to form a deep LSTM network. As shown in Figure 4; x1-x3 are the immediate input parameter content of LSTM, y1-y3 are the immediate output parameter content of LSTM, c1-c4 and h1-h4 are the long-term input parameter content and long-term output parameter content of LSTM. In the scenario of inferring the passenger's detention situation, c1 is the input for the passenger's detention time, c2 is the result of the estimated passenger's detention time, h1 is the input for the passenger's detention location, h2 is the result of the estimated passenger's detention location, x1 is the current vehicle data of the airport, and y1 is the best vehicle data estimated at the airport. c1 and h1 are inferred through LSTM, and are affected by x1, so the data of c2 and h2 can be inferred, and the corresponding x1 will also be affected by c1 to generate the data of y1. When the results are calculated, c2 and h2 become the long-term input parameters of the next LSTM unit. LSTM calculation continues. x2 can be the weather conditions of the day. c2 and h2 are affected by x2, and c3 and h3 are inferred.

S303, model training and optimization: Use historical civil aviation important event arrival and departure information data to learn and optimize the parameters of the selected LSTM model. During the training process, the prepared training set arrival and departure information data is input into the LSTM network, and then the model is trained and optimized using the validation set data. The model parameters are continuously adjusted through the back propagation algorithm and the LSTM optimizer, so that the model's loss function is gradually reduced, thereby improving the model's prediction accuracy.

S304, model evaluation and tuning: After the model training is completed, the trained model is evaluated and tuned. During the evaluation process, the data in the test set is used to evaluate the prediction performance of the model on unknown data. The evaluation indicators include mean square error, mean absolute error, and accuracy. It is used to measure the prediction accuracy and generalization ability of the model. If all three types of data are within the compliance range, the model is considered to have passed (for each type of data, the exact value of the evaluation indicator is different, and it is manually judged whether it meets the requirements). If the model performance is poor, the model can be tuned by adjusting the model structure, adjusting the hyperparameters, etc. to improve the prediction performance of the model.

Among them, mean square error (MSE): measures the degree of match between the predicted value and some true value. Its formula is as follows:

;

In the formula, is the number of samples, For real data, The predicted data represents the predicted data. In model evaluation, such as estimating the estimated arrival time of a flight based on weather conditions; In model evaluation, such as estimating the estimated arrival time of a flight based on weather conditions. Substitute multiple sets of data into the formula and calculate the final mean square error value.

Substitute multiple sets of data into the formula to find the final mean absolute error value:

;

In the formula, is the number of samples. The processing flow of the overall LSTM model is shown in Figure 5.

Information processing and prediction module 4: Apply the trained LSTM model to the new arrival and departure information data of important civil aviation activities for information processing and prediction. Through the prediction of the model, the arrival and departure of important civil aviation activities in the future can be accurately predicted, providing important decision-making support for civil aviation transportation.

Example 3: The following functions can be achieved by using the method for processing arrival and departure information of important civil aviation activities provided by the embodiment of the present invention:

Airport personnel stranded situation warning and decision-making: The flight information in recent years, the weather conditions of the routes passed by the flights, the flight departure time, arrival time, delay time, the length of time passengers leave the airport, the stranded situation of passengers at the airport, and the evacuation of stranded passengers are all input into the present invention for deep learning. After the learning is completed, the model will have a better learning situation for the situation where passengers are stranded due to flight delays. After inputting flight information, route weather conditions, and flight departure time into the model, the arrival time of the flight can be estimated, so as to make a judgment on whether the passengers will be stranded at the airport, and provide the airport terminal management department with passenger stranded warnings and passenger evacuation suggestions.

Provide event organizers with suggestions on event participants: Input the arrival and departure information of civil aviation-related passengers, the way and time of passengers going to the event venue, etc. into the model. After the model learning is completed, it can provide event organizers who hold events at the venue with predictions on the flow of people at the venue, suggestions on evacuation of people, suggestions on personnel safety risks, etc.

According to the present invention, after being implemented in an airport, it is expected that the following benefits can be obtained:

(1) The time passengers spend at the airport is reduced, and passenger congestion is reduced during peak flight hours.

(2) During the journey after passengers leave the plane, guides are arranged reasonably; the time passengers spend from leaving the plane to leaving the airport is reduced; the occupancy time of the baggage carousel is reduced, thereby reducing the power consumption of the baggage carousel; the lighting time during the passengers' journey is reduced, reducing power consumption.

(3) After leaving the airport, passengers can optimize the way to their destination. They can arrange taxis and other means of transportation in advance to avoid traffic congestion near the airport, greatly improving the utilization of nearby transportation facilities.

For events such as concerts and large conferences, arrival and departure information can be passed to other relevant departments in advance to reasonably arrange transportation and accommodation near the event venue to avoid crowds and traffic congestion.

The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principles of the present invention should be covered within the protection scope of the present invention.

Claims (7)

1. The utility model provides a civil aviation important activity supporting and separating information processing method which is characterized in that the method comprises the following steps:

s1, collecting and preprocessing a data source; the data source comprises: flight information, airport information, weather information, personnel separation information, personnel arrival intent, and nearby highway traffic information;

S2, extracting and modeling the characteristics of the data source data; the feature extraction includes: extracting indexes or attributes reflecting data characteristics and information from the original data, and inputting the data extracted by the characteristics into a resisting information prediction model established by adopting a deep learning algorithm;

s3, training and optimizing the abutment information prediction model, predicting and processing the abutment information after training and optimizing are completed, and outputting the processed result to a platform for inquiry and use;

S4, verifying and correcting the result of the processing, and correcting the result by manually intervening or adjusting the parameter of the abutting information prediction model when the abutting information prediction model predicts that the deviation or the error occurs;

S5, continuously training and optimizing the abutment information prediction model, continuously adjusting the weight proportion of the result influenced by each input parameter, and adapting to continuously changing data and requirements;

s6, putting the trained abutment information prediction model into practice, and continuously providing decision suggestions for a decision maker;

In step S2, before feature extraction, feature selection is performed, and the fields of feature selection are defined according to the frequency of occurrence of the fields and the degree of importance defined by human, where the expression is:

wherein f (n) is a characteristic value determining value, n is the occurrence frequency, ρ is a defined importance degree, ρ value is between 0 and 1, and m is the total data;

When the calculated result f (n) is larger than 0.6, the current data is considered to be characteristic, and the current numerical value is selected as a characteristic value; selecting a feature subset from the raw data that is most representative and distinguishable, the feature subset including, but not limited to:

Flight information feature: the method comprises flight departure time, arrival time, flight number and model information;

airport information features: the method comprises airport codes, terminal building and boarding gate information;

Weather information characteristics: comprises air temperature, humidity wind direction, wind speed and meteorological factors;

other information features: including the manner in which the passenger goes to the destination, follow-up planning;

in step S2, the created abutment information prediction model includes:

data set construction: dividing data collected, cleaned and extracted by features in the earlier stage into three parts, wherein one part is a training set, one part is a verification set and the other part is a test set;

Constructing a long-term memory LSTM model: using an LSTM model as a bottom layer of deep learning, selecting a plurality of LSTM layers to increase the depth of the model, and stacking the LSTM layers together in sequence to form a deep LSTM network;

In step S3, training and optimizing the abutment information prediction model includes: the LSTM training process uses a back propagation algorithm and a random gradient descent optimization algorithm; in the training process, the weight is updated by calculating the gradient of the loss function on the weight and using a random gradient descent optimization algorithm to minimize the loss function;

The prediction and processing of the abutment information after the optimization is completed after the prediction model training comprises the following steps: after the completion of the training of the resisting information prediction model, performing evaluation and tuning, wherein in the evaluation process, data in a test set are used for performing evaluation, and the prediction performance of the model on unknown data is evaluated; the evaluation index comprises a mean square error MSE, an average absolute error MAE and an accuracy; the prediction accuracy and generalization capability of the offset information prediction model are measured; if the three data are in the conforming range, the resisting information prediction model passes; if the performance of the offset information prediction model is inaccurate, tuning and optimizing are carried out by adjusting the model structure and the hyper-parameter adjustment method.

2. The civil aviation important activity withholding information processing method according to claim 1, wherein in step S1, data sources are collected, comprising: the method comprises the steps of manually uploading and extracting an interface externally provided by a third-party service system, and directly obtaining and crawling data from a database of the third-party service system; the collected data is stored in a database;

Preprocessing the data source includes:

(1) Missing value processing: identifying and processing missing values in the data, and filling the missing data; performing similar comparison and supplement filling on the missing values for the data with similar references; deleting the current data when the similar data does not exist, so as to avoid the influence of the current data on the subsequent training result;

(2) Abnormal value detection: the method comprises the steps of uniformly deleting data with obvious abnormality in the data; abnormal condition data in the data, and adjusting the data;

(3) Data format conversion: the data format is uniformly adjusted, so that the subsequent data is conveniently imported, the date and time format is uniformly changed into a standard timestamp format, all characters are changed into character formats, and the corresponding digits are reserved for the numerical values;

(4) Data deduplication: sorting the data, comparing the same kind of data, distinguishing and deleting the completely consistent data, and ensuring the uniqueness and consistency of the data;

(5) Data normalization: and (3) arranging to-be-standardized and establishing a standardized comparison library, uniformly converting the data into corresponding data in the standardized library, and conforming to uniform data standards and specifications, including but not limited to: the airline marks of all flights are adjusted to be corresponding three codes; all data time is adjusted to take Beijing time as a reference;

in step S1, after preprocessing the data source, quality evaluation is also required for the processed data.

3. The method for processing civil aviation important activity withholding information according to claim 1, wherein the mean square error MSE is used for measuring the matching degree of the predicted value and some true values, and the expression is:

Where n is the number of samples, Y _i is the real data, The predicted data is represented for the predicted data.

4. The civil aviation important activity withholding information processing method according to claim 3, wherein the average absolute error MAE is used for a sample standard deviation of a difference between a predicted value and an observed value, and the expression is:

Where m is the number of samples.

5. The civil aviation important activity withholding information processing method according to claim 1, wherein in step S4, result verification and correction are performed on the processed result, comprising: the result which is supposed to be output by the training data is calculated manually, and then the output result of the abutment information prediction model is compared with the manually calculated result.

6. A civil aviation important activity withholding information processing system, wherein the civil aviation important activity withholding information processing method according to any one of claims 1 to 5 is implemented, the system comprising:

Data acquisition and preprocessing module (1): collecting and preprocessing a data source;

Feature extraction module (2): extracting features with representativeness and distinguishing degree from the preprocessed data for training and predicting by the deep learning model;

Model training module (3): training the extracted features by using a deep learning algorithm, and constructing a resisting information prediction model;

information processing and prediction module (4): and applying the trained LSTM model to new civil aviation important activity resisting information data to process and predict information.

7. The civil aviation important activity withholding information processing system according to claim 6, wherein the feature extraction module (2) is further configured to select a feature subset having a most representative and differentiated degree from the raw data, including flight information features, airport information features, weather information features, and other information features;

the model training module (3) is used for constructing a stand-off information prediction model and comprises data set construction, long-period memory LSTM model construction, model training and optimization, and model evaluation and tuning.