A method and system for constructing a balance wheel model for civil aviation inspectors

Abstract

The present invention discloses a method and system for constructing a balance wheel model of a civil aviation inspector. Dimensional features are extracted from observable behavioral statements. A keyword matrix is ​​constructed in which rows represent dimensions and columns represent features. A three-dimensional keyword matrix is ​​constructed for civil aviation inspectors of the same level to detect the impact of different dimensions on quality competence. Based on the three-dimensional keyword matrices of different levels, the difficulty of improving quality competence at adjacent levels is detected, and a hierarchical scoring vector is constructed. Relying on information of different dimensions, a three-dimensional structure is used to map data of different dimensions to obtain the relationship between different levels and dimensions. Deep learning is used to replace the traditional manual model, so as to more accurately determine the quality competence of the corresponding inspector in different levels and dimensions.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and system for constructing a balance wheel model of a civil aviation inspector. Background Art

[0002] As aviation technology continues to advance, inspectors need to continually update their knowledge and skills to adapt to new technologies and processes. Excellent communication skills are crucial for inspectors to communicate with airlines, crew members, and other stakeholders. Inspectors also need the decision-making skills to make informed choices when faced with complex situations. Competency models also help promote the continuous improvement and development of the civil aviation regulatory system, as well as swift and effective action during emergency or accident investigations.

[0003] Determining competencies requires constructing a balance wheel model. This balance wheel model posits the competencies required of each level of inspectors, the corresponding objective (OB) for each competency, and the performance standards that should be achieved under the corresponding conditions. The research team developed the final balance wheel model based on interviews and analysis of inspectors at each level, as well as discussions and voting with senior inspectors. This method requires extensive manual judgment and wastes significant human and material resources. An alternative model is needed that accurately assesses the competencies of inspectors at multiple levels and across multiple dimensions, from new civil servants to trainee inspectors, inspectors, and senior inspectors. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for constructing a balance wheel model of a civil aviation inspector, so as to solve the above-mentioned problems existing in the prior art.

[0005] In a first aspect, an embodiment of the present invention provides a method for constructing a balance wheel model of a civil aviation inspector, comprising:

[0006] Obtain a multi-level inspector dimension structure constructed by m civil aviation inspectors; obtain n observable behavior statements corresponding to one civil aviation inspector; the multi-level inspector dimension structure indicates that civil aviation inspectors are divided into a levels;

[0007] Based on the observable behavior statements, a keyword matrix is ​​obtained; m civil aviation inspectors correspondingly obtain m keyword matrices;

[0008] Based on the multi-level inspector dimension structure and the m keyword matrices, a same-level scoring vector and a level scoring vector are obtained;

[0009] Based on observable behavioral statements, peer-level scoring and hierarchical scoring, dictionary cards are created to determine whether competencies are met.

[0010] Optionally, the step of obtaining a same-level scoring vectors and a level scoring vector based on the multi-level supervisor dimension structure and m keyword matrices includes:

[0011] According to the multi-level inspector dimension structure, the keyword matrices corresponding to the civil aviation inspectors of the same level are superimposed to obtain a three-dimensional keyword matrix; the multi-level inspector dimension structure corresponds to obtaining a three-dimensional keyword matrix;

[0012] Based on the three-dimensional keyword matrix, a peer score vector is obtained; the number of elements in the peer score vector is equal to the dimension of the n observable behavior statements;

[0013] A three-dimensional keyword matrix corresponds to a same-level scoring vector;

[0014] Based on a three-dimensional keyword matrix, a hierarchical score vector is obtained.

[0015] Optionally, obtaining the same-level score vector based on the three-dimensional keyword matrix through the first convolutional network includes:

[0016] The first convolutional network includes a two-dimensional convolution kernel with a length of 2*b; b represents the number of pages in the three-dimensional keyword matrix;

[0017] With a step size of 1, the two-dimensional convolution kernel is moved in the row direction of the three-dimensional keyword matrix for multiple convolutions to detect the relationship between observable behavior statements in one dimension and obtain the first-level keyword change feature value;

[0018] Multiple columns of the three-dimensional keyword matrix correspond to obtaining multiple first-level keyword change feature values;

[0019] Normalize the multiple first-level keyword change feature values ​​to obtain the same-level score vector.

[0020] Optionally, obtaining the hierarchical score vector based on a three-dimensional keyword matrix includes:

[0021] Through the three-dimensional detection network, the gap relationship between the three-dimensional keyword matrices of adjacent levels is detected to obtain the gap ratio;

[0022] a three-dimensional keyword matrix corresponding to a-1 gap ratios;

[0023] The divisor and the dividend of the gap ratio between adjacent levels are converted into the same value to obtain a level assignment vector; the number of elements of the level assignment vector is a.

[0024] Optionally, obtaining a keyword matrix based on the observable behavior statement includes:

[0025] Input multiple observable behavior sentences of one dimension into the sentence detection network to obtain keyword features; keyword features represent the features of multiple observable behavior sentences of one dimension;

[0026] c dimensions correspond to c keyword features;

[0027] Arrange the c keyword features in sequence to obtain the keyword matrix.

[0028] Optionally, detecting the gap relationship between three-dimensional keyword matrices at adjacent levels through a three-dimensional detection network to obtain a gap ratio includes:

[0029] Inputting the three-dimensional keyword matrix into the three-dimensional detection network, detecting the features of the keywords of the civil aviation inspectors at the same level, and obtaining the first three-dimensional keyword features;

[0030] a three-dimensional keyword matrix corresponding to a first three-dimensional keyword features;

[0031] The first three-dimensional keyword features of adjacent levels are input into the discriminant network to obtain the gap ratio.

[0032] Optionally, the training method of the three-dimensional detection network includes:

[0033] Obtain n annotation level scores; the annotation level scores represent scores for observable behavior statements at different levels;

[0034] According to the order of the observable behavior statements corresponding to the gap ratio, the difference between the two annotation level scores is calculated to obtain the annotation gap ratio;

[0035] The loss is calculated using the ratio of the labeled gaps to train the 3D detection network, and a trained 3D detection network is obtained.

[0036] Optionally, a plurality of annotated peer scores are obtained; the annotated peer scores represent scores for observable behavioral statements of civil aviation inspectors of the same level in different dimensions;

[0037] The first convolutional network is trained using the labeled same-level scores to obtain a trained first convolutional network.

[0038] Optionally, the step of establishing a dictionary card based on observable behavior statements, peer-level scoring values, and level-level scoring values ​​to determine whether competencies are qualified includes:

[0039] Construct dictionary cards based on observable behavior statements;

[0040] According to the dictionary card, n observable behavior scores are obtained; the observable behavior scores represent the scores of the n observable behaviors of the civil aviation inspector;

[0041] According to the order of the dictionary cards, construct a two-dimensional matrix of scores based on the scores of n observable behaviors;

[0042] Multiply the observable behavior scores of the same level in the score two-dimensional matrix with the corresponding same-level assigned scores, and multiply the observable behavior scores of different levels with the corresponding level assigned scores to obtain the competency matrix;

[0043] Add up multiple values ​​of one row in the multi-competency matrix to get a score vector;

[0044] The values ​​in the score vector are compared with the corresponding marked scores to determine whether the civil aviation inspector's competency is qualified.

[0045] In a second aspect, an embodiment of the present invention provides a system for constructing a balance wheel model for a civil aviation inspector, comprising:

[0046] An acquisition module is used to acquire a multi-level inspector dimension structure constructed by m civil aviation inspectors; one civil aviation inspector corresponds to acquiring n observable behavior statements; the multi-level inspector dimension structure indicates that civil aviation inspectors are divided into a levels;

[0047] A keyword matrix module is used to obtain a keyword matrix based on the observable behavior statement; m civil aviation inspectors correspondingly obtain m keyword matrices;

[0048] A scoring module, configured to obtain a same-level scoring vectors and a level scoring vector based on the multi-level inspector dimension structure and the m keyword matrices;

[0049] The discrimination module is used to establish a dictionary card based on observable behavioral statements, a same-level scoring vectors and a level scoring vector to determine whether the civil aviation inspector is qualified.

[0050] Compared with the prior art, the embodiments of the present invention achieve the following beneficial effects:

[0051] The embodiment of the present invention also provides a method and system for constructing a balance wheel model of a civil aviation inspector.

[0052] In this application, dimensional features are extracted from observable behavioral statements. Thus, a keyword matrix is ​​constructed in which rows represent dimensions and columns represent features. A three-dimensional keyword matrix is ​​constructed for civil aviation inspectors of the same level to detect the impact of different dimensions on quality competence. Based on the three-dimensional keyword matrices of different levels, the difficulty of improving the quality competence of adjacent levels is detected, and a hierarchical scoring vector is constructed. Relying on information of different dimensions, a three-dimensional structure is used to map data of different dimensions to obtain the relationship between different levels and dimensions. Deep learning is used to replace the traditional manual model, so as to achieve the technical effect of more accurately finding the quality competence of the corresponding inspector in different levels and dimensions in turn. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a flow chart of a method for constructing a balance wheel model for a civil aviation inspector provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The present invention will be described in detail below with reference to the accompanying drawings.

[0055] Example 1: Figure 1 As shown, an embodiment of the present invention provides a method for constructing a balance wheel model of a civil aviation inspector, the method comprising:

[0056] S101: Obtain a multi-level inspector dimension structure constructed by m civil aviation inspectors; obtain n observable behavior statements corresponding to one civil aviation inspector; the multi-level inspector dimension structure indicates that civil aviation inspectors are divided into a levels.

[0057] In this embodiment, the civil aviation inspectors are divided into four levels: new civil servants, trainee inspectors, inspectors, and senior inspectors, and different standards are used to judge their competence.

[0058] Among them, as in this embodiment, the observable behavior statement represents a textual statement that records the competency qualities of civil aviation inspectors in nine dimensions: political quality, job ability / technical expertise, communication and organizational coordination ability, problem-solving ability, subjective initiative, teamwork and leadership, objective and rational thinking, systematic thinking, and risk management ability.

[0059] Wherein, m, a and n are positive integers.

[0060] S102: Based on the observable behavior statements, a keyword matrix is ​​obtained; m civil aviation inspectors correspondingly obtain m keyword matrices.

[0061] S103: Based on the multi-level supervisor dimension structure and m keyword matrices, a same-level scoring vectors and a level scoring vector are obtained.

[0062] S104: Based on observable behavioral statements, peer scoring values ​​and hierarchical scoring values, a dictionary card is created to determine whether the competency is qualified.

[0063] Optionally, the step of obtaining a same-level scoring vectors and a level scoring vector based on the multi-level supervisor dimension structure and m keyword matrices includes:

[0064] According to the multi-level inspector dimension structure, the keyword matrices corresponding to civil aviation inspectors at the same level are superimposed to obtain a three-dimensional keyword matrix; the multi-level inspector dimension structure corresponds to obtaining a three-dimensional keyword matrices.

[0065] The number of rows of the three-dimensional keyword matrix is ​​equal to the number of rows of the keyword matrix, the number of columns of the three-dimensional keyword matrix is ​​equal to the number of columns of the keyword matrix, and the number of pages of the three-dimensional keyword matrix is ​​equal to the number of keyword matrices corresponding to civil aviation inspectors of the same level.

[0066] Based on the three-dimensional keyword matrix, a peer score vector is obtained; the number of elements in the peer score vector is equal to the dimension of the n observable behavior statements.

[0067] A three-dimensional keyword matrix corresponds to a same-level scoring vector;

[0068] Based on a three-dimensional keyword matrix, a hierarchical score vector is obtained.

[0069] Optionally, obtaining the same-level score vector based on the three-dimensional keyword matrix through the first convolutional network includes:

[0070] The first convolutional network includes a two-dimensional convolution kernel with a length of 2*b; b represents the number of pages in the three-dimensional keyword matrix.

[0071] Wherein, b is a positive integer.

[0072] Among them, the first convolutional network is a convolutional neural network (CNN).

[0073] With a step size of 1, the two-dimensional convolution kernel is moved in the row direction of the three-dimensional keyword matrix for multiple convolutions to detect the relationship between observable behavior statements in one dimension and obtain the first-level keyword change feature value.

[0074] Multiple columns of the three-dimensional keyword matrix correspond to obtaining multiple first-level keyword change characteristic values.

[0075] The number of first-level keyword change feature values ​​is equal to the number of dimensions, which is 9 in this embodiment.

[0076] Normalize the multiple first-level keyword change feature values ​​to obtain the same-level score vector.

[0077] In this embodiment, the normalization method is as follows: the sum of the values ​​in the keyword vectors of the same level is taken as the keyword sum, and the value in the keyword vector of the same level is divided by the keyword sum to obtain the same level score vector.

[0078] Optionally, obtaining the hierarchical score vector based on a three-dimensional keyword matrix includes:

[0079] Through the three-dimensional detection network, the gap relationship between the three-dimensional keyword matrices of adjacent levels is detected to obtain the gap ratio;

[0080] a three-dimensional keyword matrix corresponding to a-1 gap ratios;

[0081] The divisor and the dividend of the gap ratio between adjacent levels are converted into the same value to obtain a level assignment vector; the number of elements of the level assignment vector is a.

[0082] Optionally, obtaining a keyword matrix based on the observable behavior statement includes:

[0083] Input multiple observable behavior sentences of one dimension into the sentence detection network to obtain keyword features; keyword features represent the features of multiple observable behavior sentences of one dimension;

[0084] The keyword feature represents the feature of the sentence detection network when it is output without classification.

[0085] In this embodiment, the sentence detection network includes two recurrent neural networks (RNNs).

[0086] The first recurrent neural network (RNN) in the sentence detection network is used to identify keywords and is trained using the labeled keywords. The second recurrent neural network (RNN) uses these keywords to determine whether a sentence is qualified, unqualified, or excellent, and is trained using the labeled qualified, unqualified, and excellent sentences. The features before the qualified, unqualified, and excellent sentences output by the second recurrent neural network (RNN) are used as keyword features.

[0087] The c dimensions correspond to c keyword features.

[0088] Wherein, c is a positive integer;

[0089] Arrange the c keyword features in sequence to obtain the keyword matrix.

[0090] Here, multiple one-dimensional keyword features are arranged in order of fixed dimensions into a two-dimensional keyword matrix. The number of rows in the keyword matrix is ​​c, and the number of columns is equal to the length of the keyword feature.

[0091] Optionally, detecting the gap relationship between three-dimensional keyword matrices at adjacent levels through a three-dimensional detection network to obtain a gap ratio includes:

[0092] The three-dimensional keyword matrix is ​​input into the three-dimensional detection network to detect the features of the keywords of the civil aviation inspectors at the same level and obtain the first three-dimensional keyword features.

[0093] In this embodiment, the three-dimensional detection network is a three-dimensional convolutional neural network (3D Convolutional Neural Networks, 3D CNN).

[0094] The first three-dimensional keyword feature is an overall feature of keywords of multiple observable behavior statements at the same level.

[0095] a three-dimensional keyword matrix corresponding to a first three-dimensional keyword features;

[0096] Subtract the first three-dimensional keyword features of adjacent levels to obtain feature differences.

[0097] In this embodiment, the first three-dimensional keyword feature corresponding to "civil aviation inspector is a trainee inspector" is subtracted from the first three-dimensional keyword feature corresponding to "civil aviation inspector is a new civil servant" to obtain a feature difference. The first three-dimensional keyword feature corresponding to "civil aviation inspector is an inspector" is subtracted from the first three-dimensional keyword feature corresponding to "civil aviation inspector is a trainee inspector" to obtain a feature difference. The first three-dimensional keyword feature corresponding to "civil aviation inspector is an inspector" is subtracted from the first three-dimensional keyword feature corresponding to "civil aviation inspector is an inspector" to obtain a feature difference.

[0098] After the feature difference is input into the sigmoid function, it is input into the discriminant network to obtain the gap ratio.

[0099] The sigmoid function adjusts the range of the feature difference to (0, 1).

[0100] Wherein, the discriminant network adopts deep neural network (Deep Neural Networks, DNN).

[0101] Among them, the number of output neurons of the discriminant network is 1, which represents the gap ratio.

[0102] Optionally, the training method of the three-dimensional detection network includes:

[0103] Obtaining n annotated level scores; the annotated level scores represent scores of civil aviation inspectors at different levels affected by different qualities;

[0104] According to the order of the observable behavior statements corresponding to the gap ratio, the difference between the two annotation level scores is calculated to obtain the annotation gap ratio;

[0105] The loss is calculated using the ratio of the labeled gaps to train the 3D detection network, and a trained 3D detection network is obtained.

[0106] Among them, the loss is calculated by comparing the labeled gap ratio with the corresponding gap ratio through the cross entropy loss function.

[0107] Optionally, a plurality of annotated peer scores are obtained; the annotated peer scores represent scores for observable behavioral statements of civil aviation inspectors of the same level in different dimensions;

[0108] The first convolutional network is trained using the labeled same-level scores to obtain a trained first convolutional network.

[0109] Among them, the loss is calculated by combining the same-level score vector and the corresponding labeled same-level score through the cross entropy loss function.

[0110] Optionally, the step of establishing a dictionary card based on observable behavior statements, peer-level scoring values, and level-level scoring values ​​to determine whether competencies are qualified includes:

[0111] Build dictionary cards based on observable behavior statements.

[0112] The dictionary card includes conditions, performance standards and scoring standards. The conditions represent the conditions under which civil aviation inspectors make judgments. The performance standards represent the degree of mastery of professional technical knowledge and skills constituting personal behavior. The scoring standards represent the standards for scoring civil aviation inspectors of different levels.

[0113] Among them, the dictionary cards constructed by the observable behavior statements are equivalent to Case 1 and Case 2.

[0114] The dictionary card is shown in Table 1 below:

[0115] Table 1 Dictionary Cards

[0116]

[0117] Get a passing score based on the dictionary card;

[0118] According to the dictionary card, n*m observable behavior scores are obtained; the observable behavior scores represent the scores of n observable behaviors of the civil aviation inspector.

[0119] Among them, taking the qualified evaluation as the baseline, the unqualified, qualified and excellent category evaluations were assigned quantitative scores of "0, 1, 2" respectively, and then the polarization method was used to convert the quantitative evaluation scores of all levels into a percentage system.

[0120] According to the corresponding order of the dictionary cards, a two-dimensional matrix of scores is constructed based on the n*m ​​observable behavior scores.

[0121] The score two-dimensional matrix represents the corresponding observable behavior score. The number of columns in the score two-dimensional matrix is ​​n, and the number of rows is m.

[0122] Among them, in this embodiment, the columns of the score two-dimensional matrix represent nine dimensions: political quality, job ability / technical expertise, communication and organizational coordination ability, problem-solving ability, subjective initiative, teamwork and leadership, objective and rational thinking, systematic thinking and risk management ability. The score two-dimensional matrix represents civil aviation inspectors.

[0123] Multiply the observable behavior scores of the same level in the two-dimensional score matrix with the corresponding score values ​​of the same level, and multiply the observable behavior scores of different levels with the corresponding score values ​​to obtain the competency matrix.

[0124] The observable behavior scores of n dimensions corresponding to one civil aviation inspector are multiplied by the corresponding score assigned to the same level. The observable behavior scores of multiple civil aviation inspectors at different levels are multiplied by the corresponding level score to obtain the competency matrix.

[0125] Add up the multiple values ​​of one row in the multi-competency matrix to get a score vector.

[0126] The score vector represents the score of the competency of the civil aviation inspector.

[0127] The values ​​in the score vector are compared with the corresponding marked scores to determine whether the civil aviation inspector's competency is qualified.

[0128] Among them, the marking scores for each level are different. The marking scores of new civil servants are smaller than those of trainee inspectors, which are smaller than those of inspectors, which are smaller than those of senior inspectors.

[0129] Example 2: Based on the above-mentioned method for constructing a balance wheel model for a civil aviation inspector, an embodiment of the present invention further provides a system for constructing a balance wheel model for a civil aviation inspector, the system comprising:

[0130] An acquisition module is used to acquire a multi-level inspector dimension structure constructed by m civil aviation inspectors; one civil aviation inspector corresponds to acquiring n observable behavior statements; the multi-level inspector dimension structure indicates that civil aviation inspectors are divided into a levels;

[0131] A keyword matrix module is used to obtain a keyword matrix based on the observable behavior statement; m civil aviation inspectors correspondingly obtain m keyword matrices;

[0132] A scoring module, configured to obtain a same-level scoring vectors and a level scoring vector based on the multi-level inspector dimension structure and the m keyword matrices;

[0133] The discrimination module is used to establish a dictionary card based on observable behavioral statements, a same-level scoring vectors and a level scoring vector to determine whether the civil aviation inspector is qualified.

[0134] The algorithm and display provided herein are not inherently related to any particular computer, virtual system or other device. Various general-purpose systems can also be used together with the teachings based on this. According to the above description, it is obvious that the structure required for constructing this type of system. In addition, the present invention is not directed to any specific programming language. It should be understood that various programming languages ​​can be utilized to realize the content of the present invention described herein, and the above description of specific languages ​​is for the purpose of disclosing the best mode of the present invention.

[0135] In the description provided herein, numerous specific details are described. However, it is understood that embodiments of the present invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques are not shown in detail so as not to obscure the understanding of this description.

[0136] The various component embodiments of the present invention may be implemented in hardware, as software modules running on one or more processors, or as a combination thereof. Those skilled in the art will appreciate that, in practice, a microprocessor or digital signal processor (DSP) may be used to implement some or all of the functionality of some or all of the components of the apparatus according to the embodiments of the present invention. The present invention may also be implemented as an apparatus or device program (e.g., a computer program or computer program product) for performing part or all of the methods described herein. Such a program implementing the present invention may be stored on a computer-readable medium or in the form of one or more signals. Such signals may be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

Claims

1. A method for constructing a balance wheel model for a civil aviation inspector, characterized in that: include: Obtain the multi-level inspector dimension structure constructed by m civil aviation inspectors; One civil aviation inspector obtains n observable behavior statements; The multi-level inspector dimension structure indicates that civil aviation inspectors are divided into a levels; Based on the observable behavior statements, a keyword matrix is ​​obtained; m civil aviation inspectors correspondingly obtain m keyword matrices; Based on the multi-level inspector dimension structure and the m keyword matrices, a same-level scoring vector and a level scoring vector are obtained; Based on observable behavioral statements, peer-level scoring, and hierarchical scoring, dictionary cards are created to determine whether competencies are met; Based on the multi-level supervisor dimension structure and m keyword matrices, a same-level scoring vector and a level scoring vector are obtained, including: According to the multi-level inspector dimension structure, the keyword matrices corresponding to the civil aviation inspectors of the same level are superimposed to obtain a three-dimensional keyword matrix; the multi-level inspector dimension structure corresponds to obtaining a three-dimensional keyword matrix; Based on the three-dimensional keyword matrix, a peer score vector is obtained; the number of elements in the peer score vector is equal to the dimension of the n observable behavior statements; A three-dimensional keyword matrix corresponds to a same-level scoring vector; Based on a three-dimensional keyword matrix, a hierarchical scoring vector is obtained; The method of obtaining the same-level score vector based on the three-dimensional keyword matrix through the first convolutional network includes: The first convolutional network includes a two-dimensional convolution kernel with a length of 2*b; b represents the number of pages in the three-dimensional keyword matrix; With a step size of 1, the two-dimensional convolution kernel is moved in the row direction of the three-dimensional keyword matrix for multiple convolutions to detect the relationship between observable behavior statements in one dimension and obtain the first-level keyword change feature value; Multiple columns of the three-dimensional keyword matrix correspond to obtaining multiple first-level keyword change feature values; Normalize the change feature values ​​of multiple first-level keywords to obtain a same-level score vector; The hierarchical scoring vector is obtained based on a three-dimensional keyword matrix, including: Through the three-dimensional detection network, the gap relationship between the three-dimensional keyword matrices of adjacent levels is detected to obtain the gap ratio; a three-dimensional keyword matrix corresponding to a-1 gap ratios; The divisor and the dividend of the gap ratio between adjacent levels are converted into the same value to obtain a level assignment vector; the number of elements of the level assignment vector is a.

2. The method for constructing a balance wheel model for a civil aviation inspector according to claim 1, characterized in that: The keyword matrix is ​​obtained based on the observable behavior statement, including: Input multiple observable behavior sentences of one dimension into the sentence detection network to obtain keyword features; keyword features represent the features of multiple observable behavior sentences of one dimension; c dimensions correspond to c keyword features; Arrange the c keyword features in sequence to obtain the keyword matrix.

3. The method for constructing a balance wheel model for a civil aviation inspector according to claim 1, characterized in that: The method of detecting the gap relationship between three-dimensional keyword matrices at adjacent levels through a three-dimensional detection network to obtain a gap ratio includes: Input the three-dimensional keyword matrix into the three-dimensional detection network to detect the characteristics of the keywords of the civil aviation inspectors at the same level and obtain the first three-dimensional keyword features a three-dimensional keyword matrix corresponding to a first three-dimensional keyword features; The first three-dimensional keyword features of adjacent levels are input into the discriminant network to obtain the gap ratio.

4. The method for constructing a balance wheel model for a civil aviation inspector according to claim 1, characterized in that: The training method of the three-dimensional detection network includes: Obtain n annotation level scores; the annotation level scores represent scores for observable behavior statements at different levels; According to the order of the observable behavior statements corresponding to the gap ratio, the difference between the two annotation level scores is calculated to obtain the annotation gap ratio; The loss is calculated using the ratio of the labeled gaps to train the 3D detection network, and a trained 3D detection network is obtained.

5. The method for constructing a balance wheel model for a civil aviation inspector according to claim 2, characterized in that: Obtaining a plurality of annotated peer scores; the annotated peer scores represent scores for observable behavioral statements of civil aviation inspectors of the same level in different dimensions; The first convolutional network is trained using the labeled same-level scores to obtain a trained first convolutional network.

6. The method for constructing a balance wheel model for a civil aviation inspector according to claim 1, characterized in that: The above mentioned process of establishing a dictionary card based on observable behavior statements, peer-level scoring values ​​and hierarchical scoring values ​​to determine whether the competency is qualified includes: Construct dictionary cards based on observable behavior statements; According to the dictionary card, n observable behavior scores are obtained; the observable behavior scores represent the scores of the n observable behaviors of the civil aviation inspector; According to the order of the dictionary cards, construct a two-dimensional matrix of scores based on the scores of n observable behaviors; Multiply the observable behavior scores of the same level in the score two-dimensional matrix with the corresponding same-level assigned scores, and multiply the observable behavior scores of different levels with the corresponding level assigned scores to obtain the competency matrix; Add up multiple values ​​of one row in the multi-competency matrix to get a score vector; The values ​​in the score vector are compared with the corresponding marked scores to determine whether the civil aviation inspector's competency is qualified.

7. A balance wheel model construction system for civil aviation inspectors, characterized in that: include: An acquisition module is used to obtain a multi-level inspector dimension structure constructed by m civil aviation inspectors; One civil aviation inspector obtains n observable behavior statements; The multi-level inspector dimension structure indicates that civil aviation inspectors are divided into a levels; A keyword matrix module, configured to obtain a keyword matrix based on the observable behavior statement; m civil aviation inspectors correspondingly obtain m keyword matrices; A scoring module, configured to obtain a same-level scoring vectors and a level scoring vector based on the multi-level inspector dimension structure and the m keyword matrices; The discrimination module is used to build a dictionary card based on observable behavior statements, a same-level scoring vectors and a level scoring vector to determine whether the civil aviation inspector is qualified; Based on the multi-level supervisor dimension structure and m keyword matrices, a same-level scoring vector and a level scoring vector are obtained, including: According to the multi-level inspector dimension structure, the keyword matrices corresponding to the civil aviation inspectors of the same level are superimposed to obtain a three-dimensional keyword matrix; the multi-level inspector dimension structure corresponds to obtaining a three-dimensional keyword matrix; Based on the three-dimensional keyword matrix, a peer score vector is obtained; the number of elements in the peer score vector is equal to the dimension of the n observable behavior statements; A three-dimensional keyword matrix corresponds to a same-level scoring vector; Based on a three-dimensional keyword matrix, a hierarchical scoring vector is obtained; The method of obtaining the same-level score vector based on the three-dimensional keyword matrix through the first convolutional network includes: The first convolutional network includes a two-dimensional convolution kernel with a length of 2*b; b represents the number of pages in the three-dimensional keyword matrix; With a step size of 1, the two-dimensional convolution kernel is moved in the row direction of the three-dimensional keyword matrix for multiple convolutions to detect the relationship between observable behavior statements in one dimension and obtain the first-level keyword change feature value; Multiple columns of the three-dimensional keyword matrix correspond to obtaining multiple first-level keyword change feature values; Normalize the change feature values ​​of multiple first-level keywords to obtain a same-level score vector; The hierarchical scoring vector is obtained based on a three-dimensional keyword matrix, including: Through the three-dimensional detection network, the gap relationship between the three-dimensional keyword matrices of adjacent levels is detected to obtain the gap ratio; a three-dimensional keyword matrix corresponding to a-1 gap ratios; The divisor and the dividend of the gap ratio between adjacent levels are converted into the same value to obtain a level scoring vector; the number of elements of the level scoring vector is a.

Images