Air-ground application flow statistics and weight optimization management system oriented to aviation internet

Abstract

The invention relates to an air-ground application flow statistics and weight optimization management system oriented to the aviation Internet, and belongs to the technical field of the aviation Internet. According to the system, a multi-source data fusion preprocessing unit carries out standardized processing and fusion on multi-source data such as flight state, satellite network transmission, application operation and operator service quality; the traffic feature intelligent extraction unit extracts core features through feature engineering and time sequence correlation analysis and calculates the influence coefficient of the core features on the network performance; the dynamic matching decision-making unit constructs a decision-making model based on the characteristic spectrum, and deduces and generates a dynamic scheduling strategy; and the full-link optimization output unit generates an optimization report through multi-scene simulation verification and parameter adjustment. According to the method, the problems of multi-source data isomerism, feature redundancy, strategy static state and the like are solved, flow statistics precision, resource scheduling dynamic and network optimization scenario are realized, and the aviation internet resource utilization efficiency, the application response fluency and the operation stability are improved.

Description

A Traffic Statistics and Weight Optimization Management System for Air-to-Ground Applications in the Aviation Internet Technical Field

[0001] This invention belongs to the field of aviation internet technology, specifically relating to a traffic statistics and weight optimization management system for air-to-ground applications in aviation internet. Background Technology

[0002] With the rapid development of aviation internet technology, the demand for integrated air-ground network services continues to rise. Its application scenarios cover multiple fields such as flight monitoring, passenger communication, and business data transmission, which puts forward stringent requirements for network stability, resource utilization efficiency, and scenario adaptability. The core characteristics of aviation internet lie in the heterogeneity of air-to-ground links, dynamic switching between multiple operators, and real-time changes in flight status. This results in multi-dimensional and heterogeneous data sources, including aviation flight status data, satellite network transmission parameters, air-to-ground application operation data, and operator service quality indicators. These data vary in format and semantics, and are often incomplete or anomaly-prone, posing significant challenges to data integration and effective utilization. Current technologies often employ simple format conversion and splicing for multi-source data processing, lacking a standardized processing system tailored to aviation scenarios. They fail to consider differences in data importance and business logic connections, leading to low-quality fused data that struggles to support accurate decision-making. Regarding traffic feature extraction, traditional methods often focus on single-dimensional features, failing to construct a refined feature system suitable for aviation internet. This neglects the temporal patterns of traffic changes with flight phases and time periods, and fails to fully explore the deep correlation between application behavior and scenario adaptation. Furthermore, key information is easily lost during feature selection and dimensionality reduction, and the assessment of the impact of core features on network performance lacks business scenario calibration, resulting in insufficient accuracy. At the level of network resource scheduling and decision optimization, existing solutions often employ fixed weight allocation and static strategies, without dynamically adjusting based on aviation scenarios. The system failed to adequately consider constraints such as operator network capacity, application priority, and link switching frequency, and lacked adaptability design for different routes, flight phases, and network conditions. This resulted in unreasonable bandwidth allocation, inappropriate link switching timing, and difficulty in balancing resource utilization, user experience, and network stability. Furthermore, the end-to-end evaluation mechanism was incomplete, with single evaluation indicators lacking dynamic adjustment capabilities. Policy effectiveness verification relied heavily on actual operational feedback, and a simulation verification system covering all scenarios was not established. Parameter adjustments lacked scientific basis, making it difficult to achieve continuous optimization of network configuration and meet the high-quality service requirements of complex aviation internet scenarios. Summary of the Invention

[0003] To address the aforementioned problems in existing technologies, this invention provides a traffic statistics and weight optimization management system for air-to-ground applications in the aviation internet sector. The objective of this invention can be achieved through the following technical solution: It includes a multi-source data fusion preprocessing unit that acquires aviation flight status data, satellite network transmission parameters, air-to-ground application operation data, and operator service quality indicators; standardizes the format specifications and association mapping rules of various data types; integrates data resources from different sources and dimensions according to a multimodal fusion algorithm to generate a standardized fusion dataset; a traffic feature intelligent extraction unit that constructs a multi-level traffic feature dimension system based on the standardized fusion dataset; performs feature screening, dimensionality reduction, and enhancement processing using feature engineering methods; and performs time-series-based... A correlation analysis mechanism records dynamic traffic trends and extracts core features; it calculates the impact coefficient of these core features on network performance and outputs a multi-dimensional traffic feature map; a dynamic matching decision unit, based on the multi-dimensional traffic feature map, presets matching targets and constraints; it constructs a matching decision model and adds correlation weights to each influencing factor; it intelligently extrapolates and iteratively optimizes network bandwidth allocation, link switching timing, and application resource scheduling schemes to form a dynamic matching strategy scheme; a full-link optimization output unit, combined with the dynamic matching strategy scheme, presets full-process evaluation indicators; it simulates the operating effects of different strategy schemes in real-world scenarios and adjusts strategy parameters based on feedback data; based on the evaluation results and adjustment feedback, it generates a network configuration optimization report.

[0004] Specifically, the standardization process is as follows: Format specifications are established according to data types, numerical data retains the same precision, text data uses a unified encoding format, and time data uses a unified timestamp standard; a data field association mapping table is established to obtain the correspondence between fields representing the same meaning in data from different sources.

[0005] Specifically, the process of integrating data through the multimodal fusion algorithm is as follows: preprocessing each type of data separately, using the mean imputation method to supplement missing values, and removing abnormal deviation data; assigning fusion weights based on data importance, with operator service quality-related data having higher weights than basic status data; and integrating data from different dimensions into a standardized fusion dataset with a unified structure according to the association mapping rules.

[0006] Specifically, the process of constructing the multi-level traffic feature dimension system is as follows: It is divided into a basic transmission feature layer, an application behavior feature layer, and a scenario adaptation feature layer. The basic transmission feature layer covers features that directly reflect the basic performance of network transmission, and the data comes directly from the satellite network transmission parameters in the standardized fusion dataset. The application behavior feature layer, based on the basic transmission feature data and application operation data, analyzes the application's request patterns and resource consumption patterns to extract features such as application request frequency, data transmission peak, and application type proportion. The scenario adaptation feature layer integrates feature data with flight status data and operator service quality indicators to record the matching relationship between flight phases and network performance, and the matching pattern between operator networks and application types.

[0007] Specifically, the process of feature selection, dimensionality reduction, and enhancement using feature engineering methods is as follows: In the feature selection stage, a correlation analysis algorithm is used to calculate the degree of correlation between each feature and the core performance indicators of the network, while analyzing the interdependencies between features; in the feature dimensionality reduction stage, principal component analysis is used to extract the direction of data variation and retain the principal components that can reflect the core distribution pattern and key information of the data; in the feature enhancement stage, feature combination and feature transformation techniques are used to cross-combine single-dimensional features based on business logic, while optimizing the distribution pattern of features through feature transformation.

[0008] Specifically, the process of recording the dynamic trend of traffic changes and extracting core features based on the time-series correlation analysis mechanism is as follows: set a fixed time window to collect traffic data, record the peak value, average value and fluctuation range of traffic in each time window; by comparing the traffic data of different time windows, analyze the pattern of traffic changes with the flight phase and time period, and screen the core features in different scenarios.

[0009] Specifically, the process of calculating the impact coefficient of the core features on network performance is as follows: using the core features as independent variables and network performance indicators as dependent variables; establishing a mapping relationship through regression analysis algorithm to obtain the impact trend of each core feature on different performance indicators and obtain the mapping relationship result; calibrating the mapping relationship result in combination with actual business scenarios to correct the impact degree values ​​that are inconsistent with business logic, and finally outputting a coefficient that can reflect the comprehensive impact degree of each core feature on network performance.

[0010] As a preferred technical solution of the present invention, the dynamic matching decision unit takes a multi-dimensional traffic feature map as its core input. First, it combines the characteristics of dynamic switching between multiple operators and changes in flight status in aviation scenarios, and presets matching targets focusing on network resource allocation, user experience assurance, and network operation stability. It also sets constraints based on operator service quality indicators, application service importance, flight safety requirements, network carrying capacity, application priority, and link switching frequency to obtain the target priority order and constraint flexibility range under different scenarios. Then, it constructs a decision model containing an input layer, a weight calculation layer, and a strategy deduction layer. The input layer preprocesses and converts the feature data, filtering out invalid data. The weight calculation layer decomposes influencing factors using the analytic hierarchy process, compares the importance of factors pairwise to construct logical relationships, and calibrates the associated weights using historical data to ensure that the allocation fits actual needs. Finally, the strategy deduction layer uses the constraints as boundaries to perform multiple rounds of intelligent deduction according to target priority, taking into account the coordinated optimization of bandwidth allocation, link switching timing, and application resource scheduling to avoid single-dimensional decision bias. After generating multiple candidate strategies, it combines the effects of historical optimization cases to predict feasibility and optimization effects, ultimately forming a dynamic matching strategy scheme adapted to complex aviation scenarios.

[0011] Specifically, the process of setting the preset matching targets and constraints is as follows: the matching targets include network resource allocation, user experience assurance, and network operation stability; the core connotation and measurement logic of each target are obtained; the constraints include operator network carrying capacity, application priority definition, and link switching frequency control; wherein, the operator network carrying capacity is determined based on operator service quality indicators and network transmission parameter analysis; the application priority is classified according to the business importance and user dependence of the application; the link switching frequency control is set in combination with flight safety requirements and network switching stability analysis; based on business scenario requirements, the priority order of each target under different scenarios is obtained, and the elasticity range of the constraints is defined in combination with the actual operation situation.

[0012] Specifically, the process of constructing the matching decision model and adding the correlation weights of each influencing factor is as follows: it is divided into an input layer, a weight calculation layer, and a strategy deduction layer; the input layer preprocesses and converts the data based on the multi-dimensional traffic feature map; the weight calculation layer decomposes the influencing factors using the analytic hierarchy process and determines the correlation weights of each factor in combination with historical data; the strategy deduction layer optimizes step by step according to the priority of the target, with the constraints as the boundary; multiple candidate strategy schemes are generated through multiple rounds of deduction, and the feasibility and optimization effect of each scheme are predicted.

[0013] Specifically, the process of setting up full-process evaluation indicators is as follows: obtain the key evaluation dimensions of the entire process of strategy formulation, execution, and feedback, and determine the importance weight of each indicator in combination with the priority of business scenarios; based on historical operating data, industry service standards, and business demand bottom lines, set up the qualified and excellent judgment standards for each indicator and define the boundary conditions for compliance; establish a dynamic adjustment mechanism to dynamically match indicator values ​​and weight allocation according to differences in route characteristics, flight stages, and operator network status.

[0014] Specifically, the process of simulating the operational effects of different strategy schemes in real-world scenarios and adjusting strategy parameters based on feedback data is as follows: A simulation scenario library is built, containing simulation environments for different flight routes, flight phases, and operator network states; the dynamic matching strategy scheme is imported into the corresponding simulation scenario to simulate the actual operation process, and network performance indicators and application operation status data are recorded in real time; based on the deviation between the strategy parameters and the actual scenario, the parameter values ​​are adjusted in reverse.

[0015] Specifically, the process of generating the network configuration optimization report includes: network configuration parameters, policy adaptation suggestions, and performance improvement expectations; the network configuration parameters are obtained by acquiring the optimized bandwidth allocation ratio and link switching trigger conditions; the policy adaptation suggestions match corresponding policies for different application types; and the performance improvement expectations predict the improvement range of each indicator based on simulation data.

[0016] The beneficial effects of this invention are as follows: (1) By setting up a multi-source data fusion preprocessing unit and a traffic feature intelligent extraction unit, the multi-source data such as aviation flight status, satellite network transmission, air-to-ground application operation and operator service quality are first standardized. The multi-modal fusion algorithm is used to supplement missing values, remove abnormal data and assign weights according to importance. Then, feature screening, dimensionality reduction and enhancement are completed through feature engineering methods. Combined with time series correlation analysis, the core features under different scenarios are extracted and their impact coefficients on network performance are calculated. This effectively solves the problems of heterogeneous multi-source data and feature redundancy, making the traffic statistics results more accurate and the core features more in line with the needs of aviation Internet scenarios, providing high-quality support for subsequent dynamic matching decisions. (2) By setting up a dynamic matching decision unit and a full-link optimization output unit, based on the multi-dimensional traffic feature map, a decision model with integrated influence factor weights is constructed and multiple rounds of intelligent simulation are performed according to the target priority. Combined with the preset full-process evaluation indicators, the strategy effect is verified in simulation scenarios covering different routes, flight stages and operator network statuses. The parameter deviation is dynamically adjusted, and finally an optimization report containing configuration parameters, adaptation suggestions and performance expectations is generated. This enables network bandwidth allocation, link switching timing and application resource scheduling strategies to accurately adapt to various aviation scenarios, improving the resource utilization efficiency, application response smoothness and network operation stability of the aviation Internet. Attached Figure Description

[0017] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0018] Figure 1 is a system architecture diagram of a traffic statistics and weight optimization management system for air-to-ground applications oriented towards aviation internet according to the present invention; Figure 2 is a data flow diagram of a traffic statistics and weight optimization management system for air-to-ground applications oriented towards aviation internet according to the present invention; Figure 3 is a structural schematic diagram of the matching decision model in the present invention. Detailed Implementation

[0019] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0020] Please refer to Figures 1-3, which illustrate a traffic statistics and weight optimization management system for air-to-ground applications in aviation internet. This system includes: a multi-source data fusion preprocessing unit that acquires aviation flight status data, satellite network transmission parameters, air-to-ground application operation data, and operator service quality indicators; standardizes the format specifications and association mapping rules of various data types; integrates data resources from different sources and dimensions according to a multimodal fusion algorithm to generate a standardized fusion dataset; a traffic feature intelligent extraction unit that constructs a multi-level traffic feature dimension system based on the standardized fusion dataset; performs feature screening, dimensionality reduction, and enhancement processing using feature engineering methods; and records the dynamic changes in traffic based on a time-series correlation analysis mechanism. The system identifies potential and extracts core features; calculates the impact coefficient of the core features on network performance and outputs a multi-dimensional traffic feature map; a dynamic matching decision unit, based on the multi-dimensional traffic feature map, presets matching targets and constraints; constructs a matching decision model and adds correlation weights to each influencing factor; intelligently infers and iteratively optimizes network bandwidth allocation, link switching timing, and application resource scheduling schemes to form a dynamic matching strategy scheme; a full-link optimization output unit, combined with the dynamic matching strategy scheme, presets full-process evaluation indicators; simulates the operating effects of different strategy schemes in real-world scenarios and adjusts strategy parameters based on feedback data; and generates a network configuration optimization report based on the evaluation results and adjustment feedback.

[0021] Specifically, the standardization process is as follows: Format specifications are established according to data types, numerical data retains the same precision, text data uses a unified encoding format, and time data uses a unified timestamp standard; a data field association mapping table is established to obtain the correspondence between fields representing the same meaning in data from different sources.

[0022] Specifically, the process of integrating data through the multimodal fusion algorithm is as follows: preprocessing each type of data separately, using the mean imputation method to supplement missing values, and removing abnormal deviation data; assigning fusion weights based on data importance, with operator service quality-related data having higher weights than basic status data; and integrating data from different dimensions into a standardized fusion dataset with a unified structure according to the association mapping rules.

[0023] Specifically, the process of constructing the multi-level traffic feature dimension system is as follows: It is divided into a basic transmission feature layer, an application behavior feature layer, and a scenario adaptation feature layer. The basic transmission feature layer covers features that directly reflect the basic performance of network transmission, and the data comes directly from the satellite network transmission parameters in the standardized fusion dataset. The application behavior feature layer, based on the basic transmission feature data and application operation data, analyzes the application's request patterns and resource consumption patterns to extract features such as application request frequency, data transmission peak, and application type proportion. The scenario adaptation feature layer integrates feature data with flight status data and operator service quality indicators to record the matching relationship between flight phases and network performance, and the matching pattern between operator networks and application types.

[0024] In this embodiment, the multi-source data information specifically includes aviation flight status data, satellite network transmission parameters, air-to-ground application operation data, and operator service quality indicators. The aviation flight status data includes the flight stage (takeoff, level flight, landing, ground waiting), takeoff and landing status, route information, flight altitude, flight speed, flight passenger capacity, and next flight plan, directly reflecting the core data of the aviation operation scenario and providing scenario support for scenario-based traffic optimization. The satellite network transmission parameters include satellite link bandwidth, transmission delay, packet loss rate, signal strength, bit error rate, IP network activation status, and link switching response time. The air-to-ground application operation data includes application type (flight monitoring, passenger communication, service transmission, etc.), request frequency, data transmission volume, operation status (connected / disconnected / reconnected), and transmission peak value. The operator service quality indicators include network stability, IP network activation success rate, switching success rate, service availability, and fault recovery time.

[0025] In this embodiment, the core features specifically include basic transmission features (bandwidth usage, transmission latency, packet loss), application behavior features (application request frequency, data transmission peak, application type proportion), scenario adaptation features (flight phase matching degree, operator network adaptability), and key derived features that affect network performance after feature engineering screening.

[0026] In this embodiment, the influencing factors specifically include flight phase changes, operator network status (bandwidth, stability), application priority (business importance, user dependence), traffic fluctuation trends, link switching efficiency, and network carrying capacity, which are the core basis for weight allocation in the decision model; the strategy parameters specifically include bandwidth allocation ratio, link switching trigger threshold, application resource scheduling weight, strategy iteration cycle, target priority coefficient, and constraint elasticity threshold, which are the core technical parameters that can be adjusted and optimized in the dynamic matching strategy scheme.

[0027] In this embodiment, the network performance indicators specifically include bandwidth utilization and data throughput (transmission efficiency), transmission latency and request response time (response performance), packet loss rate and link switching success rate (stability), and network matching degree for flight scenarios and operator network adaptation rate (adaptability). The application operation status data specifically includes connection status and activation quantity in basic operation status, CPU / memory usage and data transmission peak in resource consumption status, request frequency and loading time in interaction performance status, and number of crashes and request timeouts in abnormal status, comprehensively covering the core monitoring dimensions of aviation internet transmission quality and air-to-ground application operation performance.

[0028] In this embodiment, the specific process of constructing a multi-level traffic feature dimension system is as follows: First, based on the business logic of air-ground collaboration of aviation internet, the system is constructed with "bottom-level support - intermediate derivation - top-level integration" as the core logic, combined with time-series data processing and correlation mining technology. When constructing the basic transmission feature layer, a sliding window sampling technique (window duration set to 1 minute, step size 30 seconds) is used to collect satellite network transmission parameters. The Kalman filter algorithm is used to smooth data fluctuations and extract basic features such as bandwidth usage, transmission delay, and packet loss. When constructing the application behavior feature layer, the Apriori association rule algorithm is used to mine the correlation between application request frequency and data transmission volume. The quartile method is used to remove outliers in the transmission peak, deriving features such as application request frequency, data transmission peak, and application type proportion. At the same time, One-Hot The encoding performs feature processing on application types; when constructing the scenario adaptation feature layer, an attention mechanism is introduced to assign weights to flight status data (such as flight altitude and stage) and operator service quality indicators (such as handover success rate), highlighting the impact of key scenario data, mining the mapping relationship between flight stage and network latency, and the adaptation rules between operator network and application type, forming features such as flight stage matching degree and operator network adaptability; finally, an inter-layer feedback mechanism is established through gradient descent method, and the sampling frequency and filtering parameters of the bottom layer features are optimized in reverse according to the evaluation results of the top layer features, ensuring that the system adapts to the dynamic changes of aviation scenarios.

[0029] Specifically, the process of feature selection, dimensionality reduction, and enhancement using feature engineering methods is as follows: In the feature selection stage, a correlation analysis algorithm is used to calculate the degree of correlation between each feature and the core performance indicators of the network, while analyzing the interdependencies between features; in the feature dimensionality reduction stage, principal component analysis is used to extract the direction of data variation and retain the principal components that can reflect the core distribution pattern and key information of the data; in the feature enhancement stage, feature combination and feature transformation techniques are used to cross-combine single-dimensional features based on business logic, while optimizing the distribution pattern of features through feature transformation.

[0030] Specifically, the process of recording the dynamic trend of traffic changes and extracting core features based on the time-series correlation analysis mechanism is as follows: set a fixed time window to collect traffic data, record the peak value, average value and fluctuation range of traffic in each time window; by comparing the traffic data of different time windows, analyze the pattern of traffic changes with the flight phase and time period, and screen the core features in different scenarios.

[0031] Specifically, the process of calculating the impact coefficient of the core features on network performance is as follows: using the core features as independent variables and network performance indicators as dependent variables; establishing a mapping relationship through regression analysis algorithm to obtain the impact trend of each core feature on different performance indicators and obtain the mapping relationship result; calibrating the mapping relationship result in combination with actual business scenarios to correct the impact degree values ​​that are inconsistent with business logic, and finally outputting a coefficient that can reflect the comprehensive impact degree of each core feature on network performance.

[0032] As a preferred technical solution of the present invention, the dynamic matching decision unit takes a multi-dimensional traffic feature map as its core input. First, it combines the characteristics of dynamic switching between multiple operators and changes in flight status in aviation scenarios, and presets matching targets focusing on network resource allocation, user experience assurance, and network operation stability. It also sets constraints based on operator service quality indicators, application service importance, flight safety requirements, network carrying capacity, application priority, and link switching frequency to obtain the target priority order and constraint flexibility range under different scenarios. Then, it constructs a decision model containing an input layer, a weight calculation layer, and a strategy deduction layer. The input layer preprocesses and converts the feature data, filtering out invalid data. The weight calculation layer decomposes influencing factors using the analytic hierarchy process, compares the importance of factors pairwise to construct logical relationships, and calibrates the associated weights using historical data to ensure that the allocation fits actual needs. Finally, the strategy deduction layer uses the constraints as boundaries to perform multiple rounds of intelligent deduction according to target priority, taking into account the coordinated optimization of bandwidth allocation, link switching timing, and application resource scheduling to avoid single-dimensional decision bias. After generating multiple candidate strategies, it combines the effects of historical optimization cases to predict feasibility and optimization effects, ultimately forming a dynamic matching strategy scheme adapted to complex aviation scenarios.

[0033] Specifically, the process of setting the preset matching targets and constraints is as follows: the matching targets include network resource allocation, user experience assurance, and network operation stability; the core connotation and measurement logic of each target are obtained; the constraints include operator network carrying capacity, application priority definition, and link switching frequency control; wherein, the operator network carrying capacity is determined based on operator service quality indicators and network transmission parameter analysis; the application priority is classified according to the business importance and user dependence of the application; the link switching frequency control is set in combination with flight safety requirements and network switching stability analysis; based on business scenario requirements, the priority order of each target under different scenarios is obtained, and the elasticity range of the constraints is defined in combination with the actual operation situation.

[0034] Specifically, the process of constructing the matching decision model and adding the correlation weights of each influencing factor is as follows: it is divided into an input layer, a weight calculation layer, and a strategy deduction layer; the input layer preprocesses and converts the data based on the multi-dimensional traffic feature map; the weight calculation layer decomposes the influencing factors using the analytic hierarchy process and determines the correlation weights of each factor in combination with historical data; the strategy deduction layer optimizes step by step according to the priority of the target, with the constraints as the boundary; multiple candidate strategy schemes are generated through multiple rounds of deduction, and the feasibility and optimization effect of each scheme are predicted.

[0035] In this embodiment, the specific process of constructing the matching decision model is as follows: A three-layer architecture of "data input - weight calculation - strategy deduction" is used, combined with data standardization and multi-objective optimization techniques to construct the model; after receiving multi-dimensional traffic feature map data, the input layer uses Z-score standardization to eliminate differences in different feature dimensions, and then uses Isolation Forest... The algorithm detects and filters invalid data (such as abnormal feature values ​​exceeding the 3σ range) to ensure the validity of the input data. The weight calculation layer uses the Analytic Hierarchy Process (AHP) as its core, decomposing the three major objectives of network resource allocation, user experience assurance, and network operation stability into primary influencing factors, which are then further subdivided into secondary factors. A judgment matrix is ​​constructed through pairwise comparisons and a consistency check is performed (CR < 0.1). At the same time, the random forest algorithm is used to train on historical optimization data to calibrate the correlation weights of each factor, so that the weight error is controlled within a preset range. The strategy inference layer uses the operator's network capacity limit and link handover frequency as constraints, and uses the particle swarm optimization (PSO) algorithm to build a multi-objective optimization framework. The optimization objectives are bandwidth utilization, response latency, and handover interruption rate. The number of iterations is set to 50, and 10 candidate strategies are generated in each iteration. Finally, the adaptation accuracy of each strategy is calculated through the confusion matrix, and cross-validation is used to avoid overfitting. The strategy with a comprehensive accuracy that meets the preset range is selected as the output, forming a matching decision model adapted to complex aviation scenarios.

[0036] Specifically, the process of setting up full-process evaluation indicators is as follows: obtain the key evaluation dimensions of the entire process of strategy formulation, execution, and feedback, and determine the importance weight of each indicator in combination with the priority of business scenarios; based on historical operating data, industry service standards, and business demand bottom lines, set up the qualified and excellent judgment standards for each indicator and define the boundary conditions for compliance; establish a dynamic adjustment mechanism to dynamically match indicator values ​​and weight allocation according to differences in route characteristics, flight stages, and operator network status.

[0037] Specifically, the process of simulating the operational effects of different strategy schemes in real-world scenarios and adjusting strategy parameters based on feedback data is as follows: A simulation scenario library is built, containing simulation environments for different flight routes, flight phases, and operator network states; the dynamic matching strategy scheme is imported into the corresponding simulation scenario to simulate the actual operation process, and network performance indicators and application operation status data are recorded in real time; based on the deviation between the strategy parameters and the actual scenario, the parameter values ​​are adjusted in reverse.

[0038] Specifically, the process of generating the network configuration optimization report includes: network configuration parameters, policy adaptation suggestions, and performance improvement expectations; the network configuration parameters are obtained by acquiring the optimized bandwidth allocation ratio and link switching trigger conditions; the policy adaptation suggestions match corresponding policies for different application types; and the performance improvement expectations predict the improvement range of each indicator based on simulation data.

[0039] In this embodiment, the scenario is set as a flight performing a cross-regional flight mission, which needs to adapt to the dynamic traffic requirements of multiple carrier satellite network switching and various air-to-ground applications such as passenger entertainment and flight monitoring. The specific process is as follows: The multi-source data fusion preprocessing unit first collects data according to the classification dimension: aviation flight status data includes the flight stage (takeoff, level flight, landing), flight altitude h1, route trajectory, etc.; satellite network transmission parameters include the link bandwidth w1 / w2, transmission delay d1 / d2, and packet loss rate l1 / l2 between carrier a and carrier b; air-to-ground application operation data includes the request frequency f1 / f2 and data transmission volume s1 / s2 of passenger video application x1 and flight monitoring application x2; carrier service quality indicators include network stability k1 / k2 and IP network opening success rate c1 / c2. Subsequently, according to the unified format specifications of data types, missing data was supplemented using the mean imputation method, and abnormal deviation data (deviation exceeding the mean ± s1 times the standard deviation) was eliminated using the 3σ principle. Fusion weights were assigned based on data importance (operator service quality data weighted q1, basic status data weighted q2, q1 > q2), and integrated into a standardized fusion dataset. The intelligent traffic feature extraction unit, based on this dataset, used sliding window sampling technology (window duration set to t1 minutes, step size t2 seconds) to collect traffic data. The Apriori association rule algorithm was used to mine the correlation between application request frequency and data transmission volume, and an attention mechanism was combined to highlight the correlation data between flight phase and network performance, constructing a multi-level traffic feature dimension system: the basic transmission feature layer extracts features such as bandwidth usage, transmission delay, and packet loss; the application behavior feature layer derives features such as the request frequency of x1 (m1 times per t1 minute) and the data transmission peak s3 of x2; the scenario adaptation feature layer forms features such as flight phase matching degree (matching degree reaches p1 in level flight phase) and operator network adaptability (operator a's adaptability to x1 is p2). After redundant features are removed by correlation analysis algorithm and dimensionality is reduced by principal component analysis method, the influence coefficient of core features on network performance is calculated by regression analysis algorithm (the influence coefficient of traffic feature x1 is y1, and x2 is y2), and a multi-dimensional traffic feature map is output. The dynamic matching decision unit takes this map as input, presets matching objectives (network resource allocation, user experience guarantee, network operation stability) and constraints (the upper limit of operator network capacity is set to p3% of bandwidth utilization, and the link switching frequency does not exceed n1 times per hour), and obtains the highest priority of user experience guarantee during the level flight phase.A three-layer decision-making model is constructed: the input layer processes data using Z-score standardization and filters invalid data using the Isolation Forest algorithm; the weight calculation layer uses the analytic hierarchy process to decompose influencing factors (flight phase, network status, application priority), and calibrates weights using historical n² flight scheduling data, with a consistency check CR < 0.1; the strategy deduction layer uses the particle swarm optimization algorithm (n³ iterations, n⁴ population size) to generate candidate strategies with constraints as boundaries, and selects strategies with a comprehensive adaptation accuracy ≥ p⁴%, determining the bandwidth allocation ratio for the level flight phase as x1 accounting for p⁵% and x2 accounting for p⁶%, and the link switching trigger threshold as switching from operator a to operator b when the latency exceeds d³ seconds; the full-link optimization output unit presets full-process evaluation indicators (covering transmission efficiency, response speed, switching stability, etc.), builds a simulation scenario library containing different routes, flight phases, and operator network statuses, imports the dynamic matching strategy into the simulation environment to simulate t³ hours of operation, and records network performance indicator data in real time. Based on feedback data, it was found that x1 had a slightly high response latency during the level flight phase. Its bandwidth allocation ratio was adjusted to p7% (p7 > p5), and the simulation was re-imported for verification. The indicator improved by p8%. Finally, a network configuration optimization report was generated to obtain the optimized bandwidth allocation logic, link switching trigger conditions, and adaptation strategies for each application, providing accurate traffic statistics and weight optimization schemes for subsequent similar flight missions of this flight.

[0040] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A traffic statistics and weight optimization management system for air-to-ground applications oriented towards aviation internet, characterized in that, include: A multi-source data fusion preprocessing unit acquires information from multiple sources; Standardize the format specifications and association mapping rules of various types of data; integrate data resources from different sources and dimensions according to multimodal fusion algorithms to generate a standardized fusion dataset; The traffic feature intelligent extraction unit constructs a multi-level traffic feature dimension system based on the standardized fusion dataset; Feature engineering methods are used for feature selection, dimensionality reduction, and enhancement. Based on the time-series correlation analysis mechanism, the dynamic trend of traffic changes is recorded and core features are extracted; Calculate the impact coefficient of the core features on network performance and output a multi-dimensional traffic feature map; the dynamic matching decision unit presets matching targets and constraints based on the multi-dimensional traffic feature map. A matching decision model is constructed, and the correlation weights of each influencing factor are added; intelligent simulation and iterative optimization are performed on network bandwidth allocation, link switching timing, and application resource scheduling schemes to form a dynamic matching strategy scheme; a full-link optimization output unit is formed, which, in combination with the dynamic matching strategy scheme, presets full-process evaluation indicators; the running effect of different strategy schemes in actual scenarios is simulated, and the strategy parameters are adjusted according to the feedback data; based on the evaluation results and adjustment feedback, a network configuration optimization report is generated.

2. The system according to claim 1, characterized in that, The specific process of standardization is as follows: Format specifications are formulated according to data types, numerical data retains the same precision, text data uses a unified encoding format, and time data uses a unified timestamp standard; a data field association mapping table is established to obtain the correspondence between fields representing the same meaning in data from different sources.

3. The system according to claim 1, characterized in that, The specific process of integrating data through the multimodal fusion algorithm is as follows: preprocessing each type of data separately, using the mean imputation method to fill in missing values, and removing abnormal deviation data; assigning fusion weights based on data importance, with operator service quality-related data having a higher weight than basic status data; Data from different dimensions are integrated into a standardized fusion dataset with a unified structure according to the association mapping rules.

4. The system according to claim 1, characterized in that, The specific process of constructing a multi-level traffic feature dimension system is as follows: It is divided into a basic transmission feature layer, an application behavior feature layer, and a scenario adaptation feature layer. The basic transmission feature layer covers features that directly reflect the basic performance of network transmission, and the data comes directly from the satellite network transmission parameters in the standardized fusion dataset. The application behavior feature layer, based on the basic transmission feature data and application operation data, analyzes the application's request patterns and resource consumption patterns to extract features such as application request frequency, data transmission peak, and application type proportion. The scenario adaptation feature layer integrates feature data with flight status data and operator service quality indicators to record the matching relationship between flight phases and network performance, and the matching pattern between operator networks and application types.

5. The system according to claim 1, characterized in that, The specific process of feature selection, dimensionality reduction and enhancement through feature engineering is as follows: In the feature selection stage, a correlation analysis algorithm is used to calculate the degree of correlation between each feature and the core indicators of network performance, and at the same time, the interdependence between features is analyzed. In the feature dimensionality reduction stage, principal component analysis is used to extract the direction of data variation and retain the principal components that can reflect the core distribution pattern and key information of the data. The feature enhancement stage employs feature combination and feature transformation techniques. Based on business logic, single-dimensional features are cross-combined, while feature transformation optimizes the distribution of features.

6. The system according to claim 1, characterized in that, The specific process of recording the dynamic trend of traffic changes and extracting core features based on the time-series correlation analysis mechanism is as follows: set a fixed time window to collect traffic data, and record the peak value, average value and fluctuation range of traffic in each time window; by comparing the traffic data of different time windows, analyze the pattern of traffic changes with the flight phase and time period, and screen the core features in different scenarios.

7. The system according to claim 1, characterized in that, The specific process for calculating the influence coefficient of the core feature on network performance is as follows: using the core feature as the independent variable and the network performance index as the dependent variable; establishing a mapping relationship through a regression analysis algorithm, obtaining the influence trend of each core feature on different performance indicators, and obtaining the mapping relationship result; The mapping results are calibrated based on actual business scenarios to correct the impact values ​​that are inconsistent with business logic, and finally output coefficients that can reflect the overall impact of each core feature on network performance.

8. The system according to claim 1, characterized in that, The specific process of setting the preset matching targets and constraints is as follows: The matching targets include network resource allocation, user experience assurance, and network operation stability; the core connotation and measurement logic of each target are obtained; the constraints include operator network carrying capacity, application priority definition, and link switching frequency control; wherein, the operator network carrying capacity is determined based on operator service quality indicators and network transmission parameter analysis; the application priority is classified according to the business importance and user dependence of the application; the link switching frequency control is set in combination with flight safety requirements and network switching stability analysis; based on business scenario requirements, the priority order of each target under different scenarios is obtained, and the elasticity range of the constraints is defined in combination with the actual operation situation.

9. The system according to claim 1, characterized in that, The specific process of constructing the matching decision model and adding the correlation weights of each influencing factor is as follows: it is divided into an input layer, a weight calculation layer, and a strategy deduction layer; the input layer preprocesses and converts the data based on the multi-dimensional traffic feature map; the weight calculation layer decomposes the influencing factors using the analytic hierarchy process and determines the correlation weights of each factor in combination with historical data; the strategy deduction layer optimizes step by step according to the priority of the target, with the constraints as the boundary; multiple candidate strategy schemes are generated through multiple rounds of deduction, and the feasibility and optimization effect of each scheme are predicted.

10. The system according to claim 1, characterized in that, The specific process of the preset full-process evaluation indicators is as follows: obtain the key evaluation dimensions of the entire process of strategy formulation, execution, and feedback, and determine the importance weight of each indicator in combination with the priority of business scenarios; Based on historical operational data, industry service standards, and business requirements, pre-set the criteria for judging the qualification and excellence of each indicator and define the boundary conditions for compliance; establish a dynamic adjustment mechanism to dynamically match indicator values ​​and weight allocation according to differences in route characteristics, flight phases, and operator network status.

11. The system according to claim 1, characterized in that, The specific process of simulating the operation effect of different strategy schemes in actual scenarios and adjusting the strategy parameters based on feedback data is as follows: build a simulation scenario library, which includes simulation environments for different routes, flight stages, and operator network states; import the dynamic matching strategy scheme into the corresponding simulation scenario, simulate the actual operation process, and record network performance indicators and application operation status data in real time; and adjust the parameter values ​​in reverse based on the deviation between the strategy parameters and the actual scenario.

12. The system according to claim 1, characterized in that, The specific process of generating the network configuration optimization report is as follows: it includes network configuration parameters, policy adaptation suggestions, and performance improvement expectations; the network configuration parameters are obtained by acquiring the optimized bandwidth allocation ratio and link switching triggering conditions; the policy adaptation suggestions are matched with corresponding policies for different application types. The performance improvement is expected based on simulation data to predict the extent of improvement for each indicator.

Images