Electronic manufacturer name identification method and system based on bill of materials

Abstract

The invention provides an electronic manufacturer name identification method and system based on a bill of materials, and aims to improve bill of materials processing efficiency, reduce manual errors and ensure product quality and cost control. The method comprises the following steps: firstly, removing irrelevant characters, spaces and special symbols in a BOM file; extracting a material feature matrix from the cleaned BOM data, performing similarity analysis on the material feature matrix and a standard electronic material feature matrix in a database, and outputting an electronic manufacturer name of which the similarity is greater than or equal to a threshold value alpha; the LSTM classification model training module is used for establishing and training an LSTM classification model; the LSTM classification model application module substitutes the material feature matrix with the similarity larger than or equal to the threshold value beta and smaller than the threshold value alpha into a final LSTM classification model, similarity calculation is carried out, and the electronic manufacturer name is output; if the feature set is smaller than the threshold value alpha, manual auditing is carried out, and the auditing electronic manufacturer name is output; the method improves the recognition efficiency, helps an enterprise optimize supply chain management, reduces the cost, and improves the efficiency of the supply chain.

Description

Technical Field

[0001] The present invention belongs to the field of bill of materials data processing. Specifically, it particularly relates to a method and system for identifying electronic manufacturer names based on a bill of materials. Background Art

[0002] In the current electronics manufacturing industry, bill of materials (BOM) management is an essential part of the production process. The BOM file details all kinds of materials required for a product and their manufacturer information, which is crucial for ensuring product quality, controlling costs, and supply chain management. However, existing bill of materials processing methods have obvious deficiencies in identifying the manufacturer names of materials. A common practice is to manually review the BOM file to extract and confirm the manufacturer name. This method is not only inefficient but also prone to errors, especially when dealing with large-scale and complex product BOMs, the problem is particularly prominent. In addition, since different manufacturers may use similar component models, but there are significant differences in quality, performance, and price, accurately identifying the manufacturer name is particularly important for ensuring product quality and cost control. Summary of the Invention

[0003] In view of the problems in the related art, the present invention proposes an electronic manufacturer name identification system based on a bill of materials to overcome the above-mentioned technical problems existing in the existing related technologies.

[0004] To solve the above technical problems, the present invention is realized through the following technical solutions:

[0005] S1. A data cleaning module, which is used to remove irrelevant characters, spaces, and special symbols in the BOM file to obtain the cleaned BOM data;

[0006] S2. A material feature matrix extraction and similarity calculation module, which is used to extract material features from the cleaned BOM data to obtain a first material feature matrix, and perform similarity analysis with the standard electronic material feature matrix in the database, and output the electronic manufacturer names corresponding to the materials in the feature set with a similarity ≥ threshold α; construct a second material feature matrix with the feature set having a similarity ≥ threshold β and < threshold α, and execute S4, and conduct manual review on the feature set with a similarity < threshold β, and output the electronic manufacturer names obtained through manual review;

[0007] S3. An LSTM classification model training module, which establishes an initial LSTM classification model, collects the feature sets of materials in other bills of materials, constructs a third material feature matrix and a fourth material feature matrix, and simultaneously collects the label matrices corresponding to the third material feature matrix and the fourth material feature matrix respectively, for training and optimizing the initial LSTM classification model to obtain the final LSTM classification model;

[0008] S4. LSTM Classification Model Application Module, which is used to substitute the second material feature matrix into the final LSTM classification model to obtain the fifth material feature matrix; calculate the similarity between the fifth material feature matrix and the standard electronic material feature matrix. If the similarity between the feature sets in the fifth material feature matrix and the feature sets in the standard electronic material feature matrix is ≥ threshold α, output the electronic manufacturer name of the corresponding material; if the similarity between the feature sets in the fifth material feature matrix and the feature sets in the standard electronic material feature matrix is < threshold α, conduct manual review and output the reviewed electronic manufacturer name.

[0009] Preferably, the S1 includes the following steps:

[0010] S11. Receive the original BOM file data;

[0011] S12. Use predefined rules to remove irrelevant characters, spaces, and special symbols from the original BOM file data and verify whether they are completely removed to obtain the BOM file data after removal; where the irrelevant characters include tab characters, line break characters, carriage return characters, etc., and the special symbols include parentheses, dashes, quotation marks, etc.

[0012] Preferably, the S12 includes the following steps:

[0013] S121. Remove all special symbols in the strings of the original BOM file data, replace multiple consecutive spaces with a single space, and remove the spaces at the beginning and end of the strings in the original BOM file data to obtain the cleaned data;

[0014] S122. Randomly extract samples from the cleaned data to check whether all irrelevant characters, spaces, and special symbols have been removed.

[0015] Preferably, the S2 includes the following steps:

[0016] S21. Extract the features of the materials from the cleaned BOM data to construct the first material feature matrix A, as follows,

[0017]

[0018] Ain represents the nth feature of the ith material in the first material feature matrix, N 1 represents a total of N 1 materials; establish the standard electronic material feature matrix E from the standard database, as follows,

[0019]

[0020] where, E in represents the nth feature of the ith electronic material in the standard electronic material feature matrix, N 2It is indicated that there are a total of N 2 electronic components; the feature sets in the first component feature matrix A and the standard electronic component feature matrix E are subjected to similarity calculation using the first similarity formula to obtain the first similarity matrix F, as follows,

[0021]

[0022] where f iN2 represents the similarity between the i-th feature set in the first component feature matrix A and the N-th 2 feature set in the standard electronic component feature matrix E;

[0023] S22. According to the first similarity matrix F, for the components corresponding to the feature sets with a similarity ≥ threshold α, directly output their electronic manufacturer names;

[0024] S23. According to the first similarity matrix F, for the feature sets with a similarity ≥ threshold β and < threshold α, construct the second component feature matrix C and execute S4, as follows,

[0025]

[0026] where C in represents the n-th feature of the i-th component in the second component feature matrix constructed from the feature sets with a similarity ≥ threshold β and < threshold α to the standard electronic component feature matrix, and N 3 indicates that there are a total of N 3 components;

[0027] S24. According to the first similarity matrix F, conduct manual review on the feature sets with a similarity < threshold β, and output the reviewed electronic manufacturer names.

[0028] Preferably, S3 includes the following steps:

[0029] S31. Collect the feature sets of the components in other component lists to form the third component feature matrix B and the fourth component feature matrix H, and the corresponding label matrices are the first label matrix B' and the second label matrix H', respectively, as follows,

[0030]

[0031] The third component feature matrix B is a training matrix, where B in represents the n-th feature of the i-th component in the training matrix, and N 4 indicates that there are a total of N 4 components in the training matrix B, and B i ′ n represents the n-th feature of the i-th component in the first label matrix; the fourth component feature matrix H is a test matrix, and Hin represents the nth feature of the ith material in the test matrix; N 5 represents that there are N materials in the training matrix H 5 materials in total, H i ′ n represents the nth feature of the ith material in the second label matrix;

[0032] S32. Establish an initial LSTM classification model, and use the training matrix B and the first label matrix B′ to train the initial LSTM classification model; after the training is completed, obtain the trained LSTM classification model, and then input the test matrix H and the second label matrix H′ into the trained LSTM classification model for testing, and optimize the trained LSTM classification model according to the test results to obtain the final LSTM classification model;

[0033] Preferably, the S32 includes the following steps:

[0034] S321. Construct an initial LSTM classification model, and set the number of iterations for training the initial LSTM classification model as a′, the batch size as b′, the optimizer as the adam optimizer, and the initial learning rate as c′;

[0035] S322. Set the current training iteration number as d and the training error threshold as d′; input the training matrix B and the first label matrix B′ into the initial LSTM classification model for training according to the batch size b′, the adam optimizer, and the initial learning rate c′. When the training error of the initial LSTM classification model < d′ or when d ≥ a′, stop the training to obtain the trained LSTM classification model;

[0036] S323. Set the accuracy threshold d″, input the test matrix H and the second label matrix H′ into the trained LSTM classification model for testing to obtain the test accuracy d″′; when d″′ ≥ d″, use the trained LSTM classification model as the final LSTM classification model; when d″′ < d″, return to S322 to continue training the trained LSTM classification model until d″′ ≥ d″ to obtain the final LSTM classification model;

[0037] By setting the relevant parameters of the LSTM classification model, then inputting the training matrix into the LSTM classification model for training to obtain the trained LSTM classification model, using the test matrix to test the trained LSTM classification model, and optimizing the trained LSTM classification model according to the test results, the LSTM classification model can be made more accurate.

[0038] Preferably, calculating the training error of the initial LSTM classification model in the S322 includes the following steps:

[0039] S3221. Substitute the training data and training result data of the initial LSTM classification model into the missing function to obtain the training error of the initial LSTM classification model. The formula of the missing function is as follows:

[0040]

[0041] where n is the number of samples, y i is the true label of the i-th sample, taking values of 0 or 1; is the predicted probability of the i-th sample, representing the probability that the model predicts that the sample belongs to class 1.

[0042] Preferably, the step S4 includes the following steps:

[0043] S41. Substitute the second material feature matrix C into the final LSTM classification model to obtain the fifth material feature matrix D, as follows:

[0044]

[0045] where D in represents the result of classifying the n-th feature of the i-th material in matrix C by the final LSTM classification model, and N 6 represents the results of classifying a total of N 6 materials in matrix C by the final LSTM classification model; use the second similarity formula to calculate the similarity between the fifth material feature matrix D and the feature set in the standard electronic material feature matrix E to obtain the second similarity matrix F'; as follows:

[0046]

[0047] where represents the similarity between the i-th feature set in the fifth material feature matrix D and the N 2 -th feature set in the standard electronic material feature matrix E; according to the second similarity matrix F', output the electronic manufacturer names corresponding to the feature sets with a similarity ≥ threshold α;

[0048] S42. According to the second similarity matrix F', for the feature sets with a similarity < threshold α, perform manual review and output the reviewed electronic manufacturer names.

[0049] An electronic manufacturer name recognition system based on a bill of materials for implementing the above-mentioned electronic manufacturer name recognition method based on a bill of materials, including a data cleaning module, a material feature matrix extraction and similarity calculation module, an LSTM classification model training module, and an LSTM classification model application module;

[0050] The data cleaning module is used to remove irrelevant content from the data and standardize the formats of the material names and manufacturer names, so as to maintain the consistency of the data;

[0051] The material feature matrix extraction and similarity calculation module is used to extract the material feature matrix and perform similarity analysis with the standard electronic material feature matrix;

[0052] The LSTM classification model training module is used to train, test and optimize the initial LSTM classification model to obtain the final LSTM classification model;

[0053] The LSTM classification model application module is used to input the originally inaccurate material feature matrix into the LSTM classification model to obtain an optimized material feature matrix; further perform similarity analysis and output the electronic manufacturer name.

[0054] The present invention has the following beneficial effects:

[0055] 1. Improve recognition efficiency: Through automated processing, the system can quickly identify and match the manufacturer names of electronic components, greatly improving the recognition efficiency and reducing the time and cost of manual review.

[0056] 2. Reduce error rate: Compared with the traditional manual review method, the method of the present invention can reduce the errors caused by manual operations. Especially when dealing with large-scale and complex product BOMs, the error rate can be significantly reduced.

[0057] 3. Accurately identifying the manufacturer name helps to ensure product quality, because components of the same model produced by different manufacturers may have significant differences in quality, performance and price.

[0058] 4. Optimize supply chain management: By accurately identifying the manufacturer name, the system can provide more accurate data support to help enterprises optimize supply chain management, reduce costs and improve the efficiency of the supply chain.

[0059] 5. Enhance system reliability: Through the combination of the iteration of the LSTM classification model and manual review, the system can more accurately identify the manufacturer names of electronic components, thereby enhancing the reliability and accuracy of the entire system.

[0060] Of course, it is not necessary for any product implementing the present invention to achieve all the above-mentioned advantages simultaneously. Brief Description of the Drawings

[0061] To more clearly illustrate the technical solutions of the embodiments of the invention, the following will briefly introduce the drawings required for describing the embodiments. Obviously, the drawings in the following description are only some embodiments of the invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0062] Figure 1 It is a schematic flowchart of a method and system for identifying the name of an electronic manufacturer based on a bill of materials according to the present invention. Detailed implementation manners

[0063] The following will clearly and completely describe the technical solutions in the embodiments of the invention with reference to the drawings in the embodiments of the invention. Obviously, the described embodiments are only some embodiments of the invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the invention without creative efforts belong to the scope of protection of the invention.

[0064] In the description of the present invention, it should be understood that the terms "opening", "upper", "lower", "top", "middle", "inner", etc. indicating the orientation or positional relationship are only for the convenience of describing the invention and simplifying the description, rather than indicating or implying that the components or elements referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation of the invention.

[0065] Embodiment 1:

[0066] Please refer to Figure 1 , the present invention discloses a method for identifying the name of an electronic manufacturer based on a bill of materials, including the following steps:

[0067] S1. A data cleaning module is used to remove irrelevant characters, spaces, and special symbols in the BOM file to obtain the cleaned BOM data;

[0068] The S1 includes the following steps:

[0069] S11. Receive the data of the original BOM file;

[0070] S12. Use predefined rules to remove irrelevant characters, spaces, and special symbols in the original BOM file data and verify whether they are completely removed to obtain the BOM file data after removal; where the irrelevant characters include tab characters, line break characters, carriage return characters, etc., and the special symbols include parentheses, dashes, quotation marks, etc.;

[0071] The S12 includes the following steps:

[0072] S121. Remove all special symbols in the strings of the original BOM file data, replace multiple consecutive spaces with a single space, and remove the spaces at the beginning and end of the strings in the original BOM file data to obtain the cleaned data;

[0073] S122. Randomly extract samples from the cleaned data to check whether all irrelevant characters, spaces, and special symbols have been removed;

[0074] S2. The material feature matrix extraction and similarity calculation module is used to extract the material features from the cleaned BOM data to obtain the first material feature matrix, and perform similarity analysis with the standard electronic material feature matrix in the database, and output the electronic manufacturer names corresponding to the feature sets with similarity ≥ threshold α; construct the second material feature matrix using the feature sets with similarity ≥ threshold β and < threshold α, and execute S4, and perform manual review on the feature sets with similarity < threshold β, and output the electronic manufacturer names obtained from the manual review;

[0075] The said S2 includes the following steps:

[0076] S21. Extract the features of the materials from the cleaned BOM data to construct the first material feature matrix A, as follows,

[0077]

[0078] Ain represents the nth feature of the ith material in the first material feature matrix, N 1 represents a total of N 1 materials; establish the standard electronic material feature matrix E from the standard database, as follows,

[0079]

[0080] where, E in represents the nth feature of the ith electronic material in the standard electronic material feature matrix, N 2 represents a total of N 2 electronic materials; calculate the similarity between the feature sets in the first material feature matrix A and the standard electronic material feature matrix E using the first similarity formula to obtain the first similarity matrix F, as follows,

[0081]

[0082] where, f iN2 represents the similarity between the ith feature set in the first material feature matrix A and the N 2 th feature set in the standard electronic material feature matrix E;

[0083] S22. According to the first similarity matrix F, for the materials corresponding to the feature sets with a similarity ≥ threshold α, directly output their electronic manufacturer names;

[0084] S23. According to the first similarity matrix F, for the feature sets with a similarity ≥ threshold β and < threshold α, construct the second material feature matrix C and execute S4 as follows,

[0085]

[0086] where C in represents the nth feature of the ith material in the second material feature matrix constructed from the feature sets with a similarity ≥ threshold β and < threshold α to the standard electronic material feature matrix, N 3 represents there are a total of N 3 materials;

[0087] S24. According to the first similarity matrix F, conduct manual review on the feature sets with a similarity < threshold β, and output the reviewed electronic manufacturer names;

[0088] S3. LSTM classification model training module, establish an initial LSTM classification model, collect the feature sets of the materials in other bills of materials, construct the third material feature matrix and the fourth material feature matrix, and at the same time collect the corresponding label matrices of the third material feature matrix and the fourth material feature matrix respectively, for training and optimizing the initial LSTM classification model to obtain the final LSTM classification model;

[0089] The S3 includes the following steps:

[0090] S31. Collect the feature sets of the materials in other bills of materials to form the third material feature matrix B and the fourth material feature matrix H, and the corresponding label matrices are the first label matrix B′ and the second label matrix H′ respectively, as follows,

[0091]

[0092] The third material feature matrix B is the training matrix, where B in represents the nth feature of the ith material in the training matrix, N 4 represents there are a total of N in the training matrix B 4 materials, B′ in represents the nth feature of the ith material in the first label matrix; the fourth material feature matrix H is the test matrix, H in represents the nth feature of the ith material in the test matrix; N 5 represents there are a total of N in the training matrix H 5 materials, H′ in represents the nth feature of the ith material in the second label matrix;

[0093] S32. Establish an initial LSTM classification model, and use the training matrix B and the first label matrix B' to train the initial LSTM classification model; after the training is completed, obtain the trained LSTM classification model, and then input the test matrix H and the second label matrix H' into the trained LSTM classification model for testing, and optimize the trained LSTM classification model according to the test results to obtain the final LSTM classification model;

[0094] The S32 includes the following steps:

[0095] S321. Construct an initial LSTM classification model, and set the number of iterations for training the initial LSTM classification model as a', the batch size as b', the optimizer as the adam optimizer, and the initial learning rate as c';

[0096] S322. Set the current training iteration number as d and the training error threshold as d'; input the training matrix B and the first label matrix B' into the initial LSTM classification model for training according to the batch size b', the adam optimizer, and the initial learning rate c', and stop the training when the training error of the initial LSTM classification model < d' or when d ≥ a' to obtain the trained LSTM classification model;

[0097] S323. Set the accuracy threshold d'', input the test matrix H and the second label matrix H' into the trained LSTM classification model for testing to obtain the test accuracy d'''; when d''' ≥ d'', use the trained LSTM classification model as the final LSTM classification model; when d''' < d'', return to S322 to continue training the trained LSTM classification model until d''' ≥ d'' to obtain the final LSTM classification model;

[0098] Calculating the training error of the initial LSTM classification model in the S322 includes the following steps:

[0099] S3221. Substitute the training data and the training result data of the initial LSTM classification model into the loss function to obtain the training error of the initial LSTM classification model; the loss function formula is as follows,

[0100]

[0101] where n is the number of samples, y i is the true label of the i-th sample, taking values 0 or 1; is the predicted probability of the i-th sample, indicating the probability that the model predicts that the sample belongs to class 1;

[0102] S4. LSTM Classification Model Application Module, which is used to substitute the second material feature matrix into the final LSTM classification model to obtain the fifth material feature matrix; calculate the similarity between the fifth material feature matrix and the standard electronic material feature matrix. If the similarity between the feature sets in the fifth material feature matrix and the feature sets in the standard electronic material feature matrix is ≥ threshold α, output the electronic manufacturer name of the corresponding material; if the similarity between the feature sets in the fifth material feature matrix and the feature sets in the standard electronic material feature matrix is < threshold α, conduct manual review and output the reviewed electronic manufacturer name.

[0103] The S4 includes the following steps:

[0104] S41. Substitute the second material feature matrix C into the final LSTM classification model to obtain the fifth material feature matrix D, as follows:

[0105]

[0106] where D in represents the result of classifying the nth feature of the ith material in matrix C by the final LSTM classification model, and N 6 represents that there are N 6 materials in matrix C that have been classified by the final LSTM classification model; use the second similarity formula to calculate the similarity between the feature sets in the fifth material feature matrix D and the standard electronic material feature matrix E to obtain the second similarity matrix F′; as follows,

[0107]

[0108] where f′ iN2 represents the similarity between the ith feature set in the fifth material feature matrix D and the Nth 2 feature set in the standard electronic material feature matrix E; according to the second similarity matrix F′, output the electronic manufacturer name of the material corresponding to the feature set with similarity ≥ threshold α.

[0109] S42. According to the second similarity matrix F′, for the feature sets with similarity < threshold α, conduct manual review and output the reviewed electronic manufacturer name.

[0110] Embodiment 2:

[0111] An electronic manufacturer name recognition system based on a bill of materials, which is used to implement the above-mentioned electronic manufacturer name recognition method based on a bill of materials, includes a data cleaning module, a material feature matrix extraction and similarity calculation module, an LSTM classification model training module, and an LSTM classification model application module;

[0112] The data cleaning module is used to remove irrelevant content from the data and standardize the formats of the material names and manufacturer names, so as to maintain the consistency of the data;

[0113] The material feature matrix extraction and similarity calculation module is used to extract the material feature matrix and perform similarity analysis with the standard electronic material feature matrix;

[0114] The LSTM classification model training module is used to train, test and optimize the initial LSTM classification model to obtain the final LSTM classification model;

[0115] The LSTM classification model application module is used to input the originally inaccurate material feature matrix into the LSTM classification model to obtain an optimized material feature matrix; further perform similarity analysis and output the electronic manufacturer name.

[0116] In the description of this specification, the descriptions referring to terms such as "one embodiment", "example", "specific example", etc. mean that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0117] The preferred embodiments of the invention disclosed above are only used to help explain the invention. The preferred embodiments do not elaborate on all the details, nor do they limit the invention to the specific embodiments described. Obviously, many modifications and variations can be made according to the content of this specification. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the invention, so that those skilled in the art in the relevant technical field can understand and utilize the invention well.

Claims

1. An electronic manufacturer name identification method based on a bill of materials, characterized in that: The following steps are involved: S1. Remove irrelevant characters, spaces and special symbols in the BOM file to obtain the cleaned BOM data; S2, extract material features from the cleaned BOM data to obtain a first material feature matrix, and perform similarity analysis with the standard electronic material feature matrix in the database, and output the electronic manufacturer name of the material corresponding to the feature set with similarity ≥ threshold α; use the feature set with similarity ≥ threshold β and < threshold α to construct a second material feature matrix, and execute S4, manually review the feature set with similarity < threshold β, and output the electronic manufacturer name manually reviewed; S3. Establish an initial LSTM classification model, collect feature sets of materials in other bills of materials, construct a third material feature matrix and a fourth material feature matrix, respectively collect label data corresponding to the third material feature matrix and the fourth material feature matrix, and train and test the initial LSTM classification model; after training and testing, obtain the final LSTM classification model; S4. Substitute the second material feature matrix into the final LSTM classification model to obtain the fifth material feature matrix; The similarity between the fifth material feature matrix and the standard electronic material feature matrix is ​​calculated. If the similarity between the feature set in the fifth material feature matrix and the feature set of the standard electronic material feature matrix is ​​≥ the threshold α, the name of the electronic manufacturer of the corresponding material is output; if the similarity between the feature set in the fifth material feature matrix and the feature set of the standard electronic material feature matrix is ​​< the threshold α, a manual review is performed and the reviewed electronic manufacturer name is output.

2. The method for identifying the name of an electronic manufacturer based on a bill of materials according to claim 1, characterized in that: The S1 comprises the following steps: S11, receiving original BOM file data; S12. Use predefined rules to remove irrelevant characters, spaces, and special symbols in the original BOM file data and verify whether they are completely removed to obtain the BOM file data after removal.

3. The method for identifying the name of an electronic manufacturer based on a bill of materials according to claim 2, characterized in that: The S12 comprises the following steps: S121, remove all special symbols in the string in the original BOM file data, replace multiple consecutive spaces with a single space, and remove the spaces at the beginning and end of the string in the original BOM file data to obtain cleaned data; S122. Randomly extract samples from the cleaned data and check whether all irrelevant characters, spaces, and special symbols have been removed.

4. The method for identifying the name of an electronic manufacturer based on a bill of materials according to claim 1, characterized in that: The S2 comprises the following steps: S21. Extract material features from the cleaned BOM data and construct the first material feature matrix A, as follows: A in represents the nth feature of the ith material in the first material feature matrix, and N1 represents a total of N1 materials; a standard electronic material feature matrix E is established from the standard database as follows: Among them, E in represents the nth feature of the ith electronic material in the standard electronic material feature matrix, and N2 represents a total of N2 electronic materials; the feature set in the first material feature matrix A and the standard electronic material feature matrix E are compared using the first similarity formula Perform similarity calculation to obtain the first similarity matrix F, as follows: Among them, f iN2 Represents the similarity between the i-th feature set in the first material feature matrix A and the N2-th feature set in the standard electronic material feature matrix E; S22. According to the first similarity matrix F, for materials corresponding to the feature set with a similarity greater than or equal to a threshold value α, directly output the name of the electronic manufacturer; S23, based on the first similarity matrix F, for the feature set with similarity ≥ threshold β and < threshold α, construct a second material feature matrix C and execute S4, as follows: Among them, C in represents the nth feature of the ith material in the second material feature matrix constructed by the feature set whose similarity with the standard electronic material feature matrix is ​​≥ threshold β and < threshold α, and N3 represents a total of N3 materials; S24. According to the first similarity matrix F, manually review the feature set with similarity less than the threshold value β, and output the reviewed electronic manufacturer name.

5. The method for identifying the name of an electronic manufacturer based on a bill of materials according to claim 1, characterized in that: The S3 comprises the following steps: S31, collect the feature sets of materials in other material lists to form the third material feature matrix B and the fourth material feature matrix H, and the corresponding label matrices are the first label matrix B' and the second label matrix H', as follows: The third material feature matrix B is a training matrix, where B in represents the nth feature of the ith material in the training matrix, N4 represents the total number of N4 materials in the training matrix B, B i ' n represents the nth feature of the i-th material in the first label matrix; the fourth material feature matrix H is the test matrix, H in represents the nth feature of the ith material in the test matrix; N5 represents that there are N5 materials in the training matrix H, i ' n represents the nth feature of the i-th material in the second label matrix; S32. Establish an initial LSTM classification model, and use the training matrix B and the first label matrix B′ to train the initial LSTM classification model; after the training is completed, a trained LSTM classification model is obtained, and then the test matrix H and the second label matrix H′ are input into the trained LSTM classification model for testing, and the trained LSTM classification model is optimized according to the test results to obtain a final LSTM classification model.

6. The method for identifying the name of an electronic manufacturer based on a bill of materials according to claim 5, characterized in that: The S32 comprises the following steps: S321, constructing an initial LSTM classification model, setting the number of iterations of the initial LSTM classification model training to a', the batch size to b', the optimizer to adam optimizer, and the initial learning rate to c'; S322, setting the current number of training iterations to d and the training error threshold to d′; inputting the training matrix B and the first label matrix B′ into the initial LSTM classification model for training according to the batch size b′, the adam optimizer and the initial learning rate c′, calculating the training error of the initial LSTM classification model, and stopping the training when the training error of the initial LSTM classification model is less than d′ or when d ≥ a′, to obtain the trained LSTM classification model; S323, set the accuracy threshold d″, input the test matrix H and the second label matrix H′ into the trained LSTM classification model for testing, and obtain the test accuracy d″′; when d″′≥d″, use the trained LSTM classification model as the final LSTM classification model; when d″′<d″, return to S322 to continue training the trained LSTM classification model until d″′≥d″, and obtain the final LSTM classification model.

7. The method for identifying the name of an electronic manufacturer based on a bill of materials according to claim 6, characterized in that: The training error of the initial LSTM classification model is calculated in S322, comprising the following steps: S3221. Substitute the training data and training result data of the initial LSTM classification model into the missing function to obtain the training error of the initial LSTM classification model. The missing function formula is as follows: Where n is the number of samples, y i is the true label of the i-th sample, which takes the value 0 or 1; is the predicted probability of the i-th sample, indicating the probability that the model predicts that the sample belongs to category 1.

8. The method for identifying the name of an electronic manufacturer based on a bill of materials according to claim 1, characterized in that: The S4 comprises the following steps: S41. Substitute the second material feature matrix C into the final LSTM classification model to obtain the fifth material feature matrix D, as follows: Among them, D in It represents the result of the nth feature of the i-th material in the matrix C after being classified by the final LSTM classification model. N6 represents the result of the total number of N6 materials in the matrix C after being classified by the final LSTM classification model. The second similarity formula is used. Calculate the similarity of the feature set in the fifth material feature matrix D and the standard electronic material feature matrix E to obtain a second similarity matrix F′; as follows, Among them, f i ' N2 Represents the similarity between the i-th feature set in the fifth material feature matrix D and the N2-th feature set in the standard electronic material feature matrix E; outputs the name of the electronic manufacturer of the material corresponding to the feature set with similarity ≥ threshold α according to the second similarity matrix F′; S42. According to the second similarity matrix F′, for the feature set with similarity less than the threshold α, manual review is performed, and the reviewed electronic manufacturer name is output.

9. An electronic manufacturer name recognition system based on a bill of materials, used to implement an electronic manufacturer name recognition method based on a bill of materials as described in any one of claims 1 to 8.

Images