Pilot flight competency assessment method based on multi-dimensional key index fusion modeling

Abstract

The invention discloses a pilot flight competency assessment method based on multi-dimensional key index fusion modeling, and relates to the technical field of computer processing, and the method comprises the following steps: S01, collecting the index data of six dimensions of physical training, clinical physical examination, aviation physiology, physiological and psychological reserve, psychological cognition and flight competency of a pilot; s02, processing the index data based on Pearson correlation analysis screening to obtain key index data, wherein the processing steps comprise: S21, carrying out standardization processing on the collected original index data; and S22, calculating the Pearson correlation coefficient r and the significance level p of each index and the pilot age. According to the method, multi-dimensional key indexes are screened based on Pearson correlation analysis, and modeling is performed by fusing four algorithms of K-means, an entropy weight method, a random forest and a neural network, so that efficient, accurate and real-time evaluation of the flight competency of the pilot is realized.

Description

Technical Field

[0001] This invention relates to the field of computer processing technology, and in particular to a method for assessing pilot flight competence based on multi-dimensional key indicator fusion modeling. Background Technology

[0002] With the rapid development of the aviation industry, flight safety has become a core concern in both civil and military aviation. Pilots, as the most critical human factor in flight missions, directly determine flight safety levels and mission execution effectiveness through their flight competence. Therefore, establishing a scientific, objective, and efficient pilot competence assessment system is of great significance for optimizing training systems, rationally allocating human resources, and preventing safety risks caused by human factors.

[0003] Currently, the assessment of pilots' capabilities both domestically and internationally mainly relies on the following methods: The assessment primarily relies on fixed-subject physical fitness tests (such as endurance and strength), written theoretical exams, and flight operation assessments (such as simulator checks and local flight tests). While these methods have been used for a long time, their evaluation results often depend on the subjective experience of the assessors and lack quantitative analysis and objective verification based on big data. Consequently, they are unable to comprehensively and accurately reflect a pilot's overall capabilities in complex and dynamic environments.

[0004] Some research or applications focus on assessing a specific dimension of a pilot's physical, psychological, or cognitive abilities. These methods overlook the fact that pilot competence is a complex system with multiple dimensions and interconnected, synergistic elements. Assessments based solely on a single dimension are one-sided and cannot provide systematic guidance for improving a pilot's overall capabilities.

[0005] In recent years, some technologies have attempted to introduce algorithmic models for assisted evaluation, but most of them use a single machine learning algorithm (such as a single neural network or decision tree). Due to the high-dimensionality, nonlinearity, and small sample size of pilot capability data, single algorithmic models often suffer from insufficient generalization ability, inadequate characterization of complex relationships, and poor stability of evaluation results, making it difficult to adapt to the diverse and demanding evaluation needs of actual flight environments.

[0006] Comprehensive analysis reveals the following major shortcomings in existing technologies: The construction of evaluation indicator systems largely relies on expert experience or simple listing, failing to utilize statistical methods (such as correlation analysis) to scientifically screen key indicators closely related to flight competence from a large pool of candidate indicators. This results in redundant evaluation dimensions, low computational efficiency, and unclear core influencing factors. Furthermore, the evaluation models depend on a single algorithm, failing to effectively integrate the advantages of different algorithms in clustering, weight determination, and classification prediction, making it difficult to meet the accuracy, robustness, and reliability requirements of high-standard applications. Summary of the Invention

[0007] To address the aforementioned technical problems, the present invention provides a method for assessing pilot flight competence based on multi-dimensional key indicator fusion modeling, the method comprising the following steps: S01. Collect indicator data from six dimensions of pilots: physical training, clinical examination, aviation physiology, physiological and psychological reserves, psychological cognition, and flight competence. S02. Based on Pearson correlation analysis, the indicator data is processed to obtain key indicator data. The processing steps include: S21. Standardize the collected raw indicator data; S22. Calculate the Pearson correlation coefficient r and its significance level p between each indicator and the pilot's age; S23. Based on the absolute value of the correlation coefficient r, |r|, and the significance level p, select the core key indicators and / or auxiliary key indicators. S03. Construct a composite evaluation model and output the pilot's flight competence fusion evaluation score and corresponding competence level based on the input key indicator data. in: The method for constructing the composite evaluation model includes: S31. Using the K-means clustering algorithm, with the objective function of minimizing the sum of squares within clusters, pilots are divided into multiple capability levels. S32. Using the entropy weight method, calculate the weight of each key indicator after screening based on the entropy value and difference coefficient of the indicator data. S33. Using the aforementioned random forest algorithm, a capability assessment and prediction model is constructed based on ensemble learning of multiple decision trees; S34. Using the neural network algorithm, construct an evaluation model that includes an input layer, a hidden layer, and an output layer; S35. Based on the accuracy of each algorithm in historical data verification, assign weights to them, and use a weighted voting method to merge the evaluation results of the four algorithms to obtain a fusion evaluation score.

[0008] Preferably, the method for screening the key indicators in step S02 includes: When |r|≥0.5 and p<0.01, the indicator is identified as a core key indicator; When 0.3 ≤ |r| < 0.5 and p < 0.05, the indicator is determined as an auxiliary key indicator; If |r| < 0.3, the index is removed.

[0009] Preferably, the objective function of the K-means clustering algorithm in step S31 is: ; in, Let i be the key indicator vector for the i-th pilot. Let be the center vector of the j-th cluster, and k be the preset number of capability levels, ranging from 3 to 5.

[0010] Preferably, in step S32, the entropy weight method is used to calculate the weight of each key indicator after screening based on the entropy value and difference coefficient of the indicator data, including... S321. Calculate the entropy value of the j-th index. : ; in, The data is the standardized data from step S21, where m is the pilot sample. S322. Calculate the difference coefficient of the j-th indicator. : ; S323. Calculate the weight of the j-th indicator. : ; Where n is the number of key indicators.

[0011] Preferably, in step S33, the random forest algorithm is used to construct a capability assessment prediction model based on the ensemble learning of multiple decision trees. This includes constructing a training set for each decision tree based on Bootstrap sampling and splitting nodes using random feature selection, wherein feature importance is measured based on the Gini coefficient.

[0012] Preferably, the weighted voting method used in step S35 to fuse the evaluation results of the four algorithms satisfies the following formula: ; in, , , , The weights of the four algorithms are respectively, satisfying... + + + =1.

[0013] Preferably, when performing step S03 to output the pilot's flight competence fusion assessment score and corresponding competence level, at least the following two assessment modes are available for the user to choose from, including: Fast evaluation mode: Evaluation is performed using only the aforementioned random forest algorithm, with an evaluation time of less than 10 seconds; Precise evaluation mode: The evaluation is performed by fusing the four algorithms mentioned above, and the evaluation time is less than 60 seconds.

[0014] Preferably, the corresponding capability in step S03 includes pre-classifying the pilot's flight competence into four levels: excellent, good, qualified, and unqualified based on the fusion evaluation score threshold.

[0015] The present invention has at least the following beneficial effects: 1. By introducing Pearson correlation analysis to statistically screen a large number of initial indicators, key indicators that are strongly or moderately correlated with pilots' age are automatically identified, effectively eliminating redundant and weakly correlated indicators.

[0016] 2. It innovatively integrates four algorithms—K-means clustering, entropy weighting, random forest, and neural network—for fusion modeling, overcoming the limitations of single-algorithm models, such as poor generalization ability and susceptibility to data features. It combines the advantages of different algorithms through a weighted voting mechanism. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of the indicator data acquisition process provided in Embodiment 1 of the present invention; Figure 2 A flowchart provided for Embodiment 1 of the present invention; Figure 3 This is a correlation analysis chart between various dimensional indicators and age provided in Embodiment 1 of the present invention; Figure 4 This is a comparative analysis chart of the multi-algorithm fusion evaluation results provided in Embodiment 1 of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "including," "having," and any variations are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0021] Example 1 This embodiment provides a method for assessing pilot flight competence based on multi-dimensional key indicator fusion modeling. The method includes the following steps: Figure 2 As shown: S01. Collect indicator data from six dimensions of pilots: physical training, clinical examination, aviation physiology, physiological and psychological reserves, psychological cognition, and flight competence.

[0022] Specifically, such as Figure 1 As shown, it connects to various preset testing devices, such as physical fitness testing instruments (e.g., electronic timing track, strength testing instrument), medical equipment (e.g., electrocardiograph, blood pressure monitor), aviation physiological simulators (e.g., centrifuge, hypobaric chamber), and psychological cognitive testing terminals (e.g., touch tablet with dedicated software).

[0023] S02. Based on Pearson correlation analysis, the indicator data is processed to obtain key indicator data. The processing steps include: S21. Standardize the collected raw indicator data; S22. Calculate the Pearson correlation coefficient r and its significance level p between each indicator and the pilot's age; S23. Based on the absolute value of the correlation coefficient r, |r|, and the significance level p, select the core key indicators and / or auxiliary key indicators.

[0024] Specifically, such as Figure 3 The preprocessing server automatically initiates the standardization process after receiving the indicator data transmitted by S01. Its core operation involves calling the global mean (μ) and standard deviation (σ) of each indicator, pre-stored in the database and calculated during the historical data training phase. For each raw data value Xij (corresponding to the j-th indicator for the i-th pilot), the system automatically applies the formula: =(X ij -μ j ) / σ jThe system performs calculations. This process eliminates the influence of differences in units (such as seconds, millivolts, and times) and orders of magnitude between different indicators, converting all indicators into a distribution with a mean of 0 and a standard deviation of 1. For rare new indicators that cannot be automatically matched or data with incorrect formatting, the system will trigger an exception handling procedure, such as prompting the operator for confirmation or using preset default value logic.

[0025] Furthermore, the method for selecting key indicators in step S02 above includes: When |r|≥0.5 and p<0.01, the indicator is identified as a core key indicator; When 0.3 ≤ |r| < 0.5 and p < 0.05, the indicator is determined as an auxiliary key indicator; If |r| < 0.3, the index is removed.

[0026] Specifically, for each preprocessed indicator to be screened (e.g., 3000-meter run time, heart rate variability), the following operations are performed: First, a statistical computing library (such as Python's SciPy) is called to calculate the Pearson correlation coefficient r and its significance p-value between the indicator's data sequence and the corresponding pilot's age sequence. Then, the system automatically compares the calculated |r| and p-value with preset thresholds and classifies the indicator based on the comparison results. For example, if an indicator has |r|=0.58 and p=0.002, the system determines that it meets the condition of |r|≥0.5 and p<0.01, automatically marks it, and stores it in the core key indicator set. Finally, the system outputs the filtered key indicator set for subsequent modeling, while weakly correlated indicators are excluded from subsequent processes.

[0027] S03. Construct a composite evaluation model and output the pilot's flight competence fusion evaluation score and corresponding competence level from the input key indicator data. in: Methods for constructing composite evaluation models include: S31. Using the K-means clustering algorithm, with the objective function of minimizing the sum of squares within clusters, pilots are divided into multiple capability levels. The objective function of the K-means clustering algorithm in this embodiment is: ; in, Let i be the key indicator vector for the i-th pilot. Let be the center vector of the j-th cluster, and k be the preset number of capability levels, ranging from 3 to 5.

[0028] Specifically, the K-means algorithm is run using a historical dataset of pilots' key performance indicators (KPIs). First, the number of clusters k (e.g., k=4) is determined based on operational requirements (typically corresponding to levels such as excellent, good, satisfactory, under observation, and unsatisfactory). After initializing k cluster centers, the algorithm enters an iterative process: calculating the Euclidean distance from each pilot's data point (i.e., their KPI vector) to all cluster centers and assigning it to the nearest cluster; subsequently, the mean of all data points in each cluster is recalculated and updated as the new center of that cluster. This process is repeated until the cluster centers no longer change significantly or the maximum number of iterations is reached, the objective function J (the sum of squared distances from all data points to their respective cluster centers) tends to a minimum, and the algorithm converges. Finally, the algorithm outputs k clusters and the center vector of each cluster. These clusters do not inherently carry labels of excellent or satisfactory; domain experts need to interpret and name each cluster post-hocly based on the known performance of pilots within the cluster (such as accident records, instructor evaluations, and mission completion rates), thus mapping the mathematical clusters 1, 2, ... to operational competence levels A, B, ... Subsequently, these k cluster center vectors were fixed and used as a reference system for assessing the ability level of new pilots.

[0029] S32. Using the entropy weight method, based on the entropy value and difference coefficient of the indicator data, calculate the weight of each key indicator after screening, including... S321. Calculate the entropy value of the j-th index. : ; in, The data is the standardized data from step S21, where m represents the pilot sample. S322. Calculate the difference coefficient of the j-th indicator. : ; S323. Calculate the weight of the j-th indicator. : ; Where n is the number of key indicators.

[0030] Specifically, the input is a standardized data matrix of all pilots to be evaluated (e.g., a batch of 100) across various key indicators. The program strictly follows the three-step formula for calculation: First, it calculates the proportion (p0.05) of all pilot data for each indicator relative to the total for that indicator. ij This reflects the relative performance of a single sample on this indicator; secondly, the entropy value (e) of this indicator is calculated based on the proportion. j The smaller the entropy value, the greater the data variability and the richer the information for that indicator; then, the coefficient of variation is calculated. The entropy value is transformed into a positive information content measure; finally, the difference coefficients of all indicators are summed and normalized to obtain the final weight of each indicator (w).j This ensures that the sum of all weights is 1. The calculated weight vector will be directly used in subsequent weighted score calculations (Score_entropy=Σw). j *x j ) S33. Using the random forest algorithm, an ability assessment prediction model is constructed based on the ensemble learning of multiple decision trees. Specifically, the ability assessment prediction model is constructed based on the ensemble learning of multiple decision trees. Each decision tree is used to construct a training set based on Bootstrap sampling, and node splitting is performed using random feature selection. The importance of features is measured based on the Gini coefficient.

[0031] Specifically, a random forest model is constructed using historical pilot data (including key indicators and known capability level labels).

[0032] First, Bootstrap sampling is performed: Assuming there are 1000 historical data points, the system will perform 100 random samplings with replacement (corresponding to 100 decision trees). Each sampling will extract approximately 630 data points as the training set for a tree, while the approximately 370 data points that are not selected will constitute the out-of-bag data for that tree, used for internal validation.

[0033] Next, when splitting at each node of each decision tree, random feature selection is used. For example, if the current node needs to judge 20 key indicators, the system will not consider all 20, but will randomly select a subset (e.g., √20 ≈ 4 indicators) as candidates, and choose the best split point from among them. This best judgment criterion is based on the Gini coefficient: the system calculates the degree of improvement in the purity of pilot ability level in the left and right child nodes after splitting each candidate feature. The lower the Gini coefficient, the purer the category (e.g., excellent, qualified) of the samples in the node. Suppose that at a certain node, from four randomly selected features—anti-G ability, 3000-meter run time, attention duration, and heart rate variability—it is found that splitting according to anti-G ability ≥ the standard value can minimize the Gini coefficient to the greatest extent (i.e., allowing more excellent pilots to be assigned to the left subtree and more qualified pilots to be assigned to the right subtree), then this node will choose anti-G ability as the split feature. In this way, the system trains 100 decision trees with different structures and based on different data subsets and feature perspectives, forming a forest.

[0034] When a new pilot needs to be evaluated, the system simultaneously inputs their standardized key performance indicator data (such as G-force tolerance 0.85, 3000-meter run 0.72, etc.) into these 100 pre-trained decision trees. Each tree then proceeds as if traversing a process. Figure 1Similarly, based on its own series of node splitting rules (e.g., the first tree might first determine that the resilience is >0.8, and the second tree might first determine that the attention duration is >0.7), the pilot is ultimately guided to a leaf node, which corresponds to a capability score derived from historical data (e.g., falling into a leaf node with a cluster of excellent trees results in an output score of 92). Finally, the system collects the outputs of all trees, assigns weights based on the accuracy of each tree's out-of-bag data validation during training, and performs weighted voting or averaging. For example, if 60 out of 100 trees give an excellent score (score >85), the final weighted average score is 88. Simultaneously, the system also calculates the frequency with which resilience is used as a key splitting feature across all trees and the total decrease in the Gini coefficient resulting from it, using this as a quantitative output of the feature importance of this indicator, providing interpretive evidence for the evaluation report (e.g., resilience contributes 18% in this round of evaluation).

[0035] S34. Using a neural network algorithm, construct an evaluation model that includes an input layer, a hidden layer, and an output layer; The system interface provides evaluators with two preset mode options: Quick Evaluation (using only Random Forest) and Precise Evaluation (integrating four algorithms). The evaluator's selection is passed to the backend system as a key parameter. The system then dynamically triggers the corresponding computational pipeline based on the received mode instruction. For example, if Precise Evaluation is selected, the system loads pre-trained models of K-means, Entropy Weights, Random Forest, and Neural Network in parallel. Through an internal task scheduler, the pre-processed pilot key indicator vectors are simultaneously distributed to these four algorithm computation units. Each unit operates independently: K-means calculates its distance to the center of the historical capability cluster to determine the baseline score; Entropy Weights dynamically calculates indicator weights based on the current data batch and performs a weighted sum; Random Forest allows hundreds of decision trees to vote; and Neural Network derives predictions through forward propagation. After all algorithms complete their computations within the specified parallel time window, the system calculates a weighted average of the output scores of each algorithm using preset fusion weights determined based on historical validation accuracy (e.g., K-means: 0.2, entropy weight method: 0.2, random forest: 0.3, neural network: 0.3). Finally, the system maps this score to a specific capability level according to a preset threshold range (e.g., excellent: 85-100 points), thus completing the evaluation.

[0036] S35. Based on the accuracy of each algorithm in historical data verification, assign weights to them, and use a weighted voting method to merge the evaluation results of the four algorithms to obtain a fusion evaluation score.

[0037] In the above embodiments, the weighted voting method used to fuse the evaluation results of the four algorithms satisfies the following formula: ; in, , , , The weights of the four algorithms are respectively, satisfying... + + + =1.

[0038] Specifically, such as Figure 4 As shown, a labeled historical pilot dataset (i.e., samples whose true ability levels are known) is used to evaluate the pilots using four pre-trained algorithms. The evaluation scores of each algorithm are compared with the true labels to calculate the historical accuracy of each algorithm (e.g., random forest has an accuracy of 94%, and neural network has 93%). After normalizing the accuracy (making the sum equal to 1), the resulting weights are determined as fixed weights for each algorithm (e.g., w). rf =0.3, w nn =0.3, w kmeans =0.2, w entropy =0.2). This process ensures that the algorithm with the better historical performance has a greater say in the final decision. Subsequently, in the real-time evaluation phase, when calculating for the pilots being evaluated, the system runs four algorithms in parallel, each producing four raw scores. These four scores are then weighted and summed with predetermined weights to obtain a unique fusion score, which is used to classify the capability level. The scoring criteria are as follows: 85-100 is excellent, 70-84 is good, 69-50 is satisfactory, and below 50 is unsatisfactory.

[0039] Example 2 This invention provides a non-transitory computer-readable storage medium storing at least one instruction or at least one program segment, which is loaded and executed by a processor to implement the following steps: Data on pilots' physical training, clinical examination, aviation physiology, physiological and psychological reserves, psychological cognition, and flight competence were collected. The key indicator data is obtained by screening and processing the indicator data based on Pearson correlation analysis. A composite assessment model is constructed, and the key indicator data input is used to output the pilot's flight competence integrated assessment score and corresponding competence level.

[0040] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0041] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0042] Example 3 This invention provides an electronic device, including a processor and a memory, wherein the memory stores at least one instruction or at least one program segment, and the at least one instruction or the at least one program segment is loaded and executed by the processor to implement the following steps: Data on pilots' physical training, clinical examination, aviation physiology, physiological and psychological reserves, psychological cognition, and flight competence were collected. The key indicator data is obtained by screening and processing the indicator data based on Pearson correlation analysis. A composite assessment model is constructed, and the key indicator data input is used to output the pilot's flight competence integrated assessment score and corresponding competence level.

[0043] 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 method for assessing pilot flight competence based on multi-dimensional key indicator fusion modeling, characterized in that, The method includes the following steps: S01. Collect indicator data from six dimensions of pilots: physical training, clinical examination, aviation physiology, physiological and psychological reserves, psychological cognition, and flight competence. S02. Based on Pearson correlation analysis, the indicator data is processed to obtain key indicator data. The processing steps include: S21. Standardize the collected raw indicator data; S22. Calculate the Pearson correlation coefficient r and its significance level p between each indicator and the pilot's age; S23. Based on the absolute value of the correlation coefficient r, |r|, and the significance level p, select the core key indicators and / or auxiliary key indicators. S03. Construct a composite evaluation model and output the pilot's flight competence fusion evaluation score and corresponding competence level based on the input key indicator data. in: The method for constructing the composite evaluation model includes: S31. Using the K-means clustering algorithm, with the objective function of minimizing the sum of squares within clusters, pilots are divided into multiple capability levels. S32. Using the entropy weight method, calculate the weight of each key indicator after screening based on the entropy value and difference coefficient of the indicator data. S33. Using the aforementioned random forest algorithm, a capability assessment and prediction model is constructed based on ensemble learning of multiple decision trees; S34. Using the neural network algorithm, construct an evaluation model that includes an input layer, a hidden layer, and an output layer; S35. Based on the accuracy of each algorithm in historical data verification, assign weights to them, and use a weighted voting method to merge the evaluation results of the four algorithms to obtain a fusion evaluation score.

2. The method for assessing pilot flight competence based on multi-dimensional key indicator fusion modeling as described in claim 1, characterized in that, The method for screening the key indicators in step S02 includes: When |r|≥0.5 and p<0.01, the indicator is identified as a core key indicator; When 0.3 ≤ |r| < 0.5 and p < 0.05, the indicator is determined as an auxiliary key indicator; If |r| < 0.3, the index is removed.

3. The method for assessing pilot flight competence based on multi-dimensional key indicator fusion modeling as described in claim 1, characterized in that, The objective function of the K-means clustering algorithm in step S31 is: ； in, Let i be the key indicator vector for the i-th pilot. Let be the center vector of the j-th cluster, and k be the preset number of capability levels, ranging from 3 to 5.

4. The method for assessing pilot flight competence based on multi-dimensional key indicator fusion modeling as described in claim 1, characterized in that, In step S32, the entropy weight method is used to calculate the weight of each key indicator after screening, based on the entropy value and difference coefficient of the indicator data. S321. Calculate the entropy value of the j-th index. : ; in, The data is the standardized data from step S21, where m is the pilot sample. S322. Calculate the difference coefficient of the j-th indicator. : ; S323. Calculate the weight of the j-th indicator. : ; Where n is the number of key indicators.

5. The method for assessing pilot flight competence based on multi-dimensional key indicator fusion modeling as described in claim 1, characterized in that, In step S33, the random forest algorithm is used to construct a capability assessment prediction model based on the ensemble learning of multiple decision trees. This includes constructing a training set for each decision tree based on Bootstrap sampling and splitting nodes using random feature selection. The importance of features is measured based on the Gini coefficient.

6. The method for assessing pilot flight competence based on multi-dimensional key indicator fusion modeling as described in claim 1, characterized in that, In step S35, the weighted voting method used to fuse the evaluation results of the four algorithms satisfies the following formula: ； in, , , , The weights of the four algorithms are respectively, satisfying... + + + =1.

7. The method for assessing pilot flight competence based on multi-dimensional key indicator fusion modeling as described in claim 1, characterized in that, When performing step S03, which outputs the pilot's flight competence fusion assessment score and corresponding competence level, at least the following two assessment modes are available for the user to choose from: Fast evaluation mode: Evaluation is performed using only the aforementioned random forest algorithm, with an evaluation time of less than 10 seconds; Precise evaluation mode: The evaluation is performed by fusing the four algorithms mentioned above, and the evaluation time is less than 60 seconds.

8. The method for assessing pilot flight competence based on multi-dimensional key indicator fusion modeling as described in claim 1, characterized in that, The corresponding capability in step S03 includes pre-classifying the pilot's flight competence into four levels: excellent, good, qualified, and unqualified based on the fusion evaluation score threshold.

9. A non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores at least one instruction or at least one program segment, characterized in that, The at least one instruction or the at least one program segment is loaded and executed by the processor to implement the steps of the pilot flight competence assessment method based on multi-dimensional key indicator fusion modeling as described in any one of claims 1-8.

10. An electronic device, characterized in that, The method includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, the at least one instruction or the at least one program being loaded and executed by the processor to implement the steps of the pilot flight competence assessment method based on multi-dimensional key indicator fusion modeling as described in any one of claims 1-8.

Images