A neural network-based method for extracting building materials equipment identification in industrial Internet
Through the correlation analysis model based on neural network, the problem of extracting identifiers from ordinary character strings in industrial Internet identification parsing is solved, the recognition accuracy and speed are improved, and the efficient extraction of building materials equipment identification is achieved.
Patent Information
- Application Number
- CN202410645274.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-05-23
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-05-23
AI Technical Summary
In the resolution of industrial Internet identification, there are challenges in extracting industrial Internet identification from ordinary strings, especially due to the customization of the identification length and the lack of general rules, resulting in low recognition accuracy and speed.
Using a neural network-based method, a correlation analysis model is constructed, and the correlation between the target iconic character segment and the industrial Internet identity is analyzed by training the neural network system. The recurrent neural network, convolutional neural network and attention mechanism are used to optimize Model parameters to improve identification accuracy and speed.
It effectively improves the accuracy of identifying building materials equipment identification from ordinary label strings and the speed of extracting identification information, ensuring the accurate analysis of the identification content.
Smart Images

Figure CN118445456B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of information extraction, and in particular to a method for extracting identification of building materials equipment in the industrial Internet based on a neural network. Background Art
[0002] The industrial Internet identification resolution system is an important part of the industrial Internet network system and the nerve center that supports the interconnection of the industrial Internet. It plays a vital role in the development of the industrial Internet. The industrial Internet identification resolution platform connects all production factors such as equipment, machines, and materials in the industry by establishing a unified identification system. It connects fragmented data and applications through the resolution system, and realizes the mastery of information such as data sources, flow processes, and usage trends, which can realize the full life cycle management of the entire industrialized production.
[0003] The core of the entire industrial Internet identification resolution system includes identification encoding, identification resolution system and identification data service. Before identification resolution, how to extract the industrial Internet identification from ordinary strings is a particularly important issue. It is easy to extract an ordinary string with a known prefix and length. However, the industrial Internet identification uses different prefixes to indicate different regions, different companies, etc., and the identification length is customized by the identifier. There is no universal rule to follow. These problems have brought challenges to the extraction of industrial Internet identification from ordinary strings.
[0004] Based on the above problems, the present invention proposes a neural network-based industrial Internet building material equipment identification extraction method, which effectively achieves the purpose of extracting building material equipment identification from ordinary character strings. Summary of the invention
[0005] The present invention provides a method for extracting industrial Internet building materials equipment identification based on a neural network, which is used to build an association analysis model by training a neural network to analyze the correlation between a target iconic character segment selected from a common character string containing the industrial Internet identification of building materials equipment and the industrial Internet identification, and parse the analysis result to obtain the content of the identification, thereby effectively improving the accuracy of identifying the building materials equipment identification from the common label character string and the speed of extracting the identification information.
[0006] The present invention provides a method for extracting identification of building materials equipment in the industrial Internet based on a neural network, comprising:
[0007] Step 1: Obtain a common string containing the building materials equipment industrial Internet logo, and select several target iconic character segments from it;
[0008] Step 2: Analyze the correlation between the target iconic character segment and the industrial Internet logo using the association analysis model obtained by training the neural network system based on the effective logo data in the industrial Internet platform to obtain the initial search results;
[0009] Step 3: Obtain and parse the final search results based on the initial search results to obtain the identified content.
[0010] Preferably, a common character string containing the building materials equipment industrial Internet logo is obtained, and several target iconic character segments are selected therefrom, including:
[0011] Get a common string containing the industrial Internet identifier of building materials equipment;
[0012] According to the classification of the building materials equipment industrial Internet identification information, several character segments are screened from the obtained common character strings and output as target iconic character segments.
[0013] Preferably, the industrial Internet identifier is created by an industrial Internet identifier resolution secondary node platform.
[0014] Preferably, the correlation analysis model obtained by training the neural network system based on the effective identification data in the industrial Internet platform is used to analyze the correlation between the target iconic character segment and the industrial Internet building materials equipment identification to obtain several initial search results, including:
[0015] Performing data preprocessing on valid identification data obtained from the industrial Internet platform to obtain a first data set;
[0016] Divide the first data set according to a preset division ratio to obtain training data and test data;
[0017] After adding an additive attention mechanism and a CBAM attention mechanism to the recurrent neural network and the convolutional neural network respectively, the training data is used to train the neural network to obtain an initial analysis model;
[0018] Inputting the test data into the initial analysis model to obtain test analysis results;
[0019] Based on the test analysis results, determining a model performance index of the initial analysis model;
[0020] When the model performance index is not greater than a set performance threshold, optimizing the model parameters of the initial analysis model to obtain a correlation analysis model;
[0021] Otherwise, outputting the initial analysis model as a correlation analysis model;
[0022] The target iconic character segment is input into the association analysis model to obtain the correlation between the target iconic character segment and the industrial Internet building materials equipment identification, and output it as the initial search result.
[0023] Preferably, based on the test analysis results, determining the model performance index of the initial analysis model includes:
[0024] Extract historical target feature data of current test data within a preset historical time period and the historical correlation with the industrial Internet identifier from the identifier extraction database;
[0025] Determine the historical target characteristic value based on the historical target characteristic data, and construct the characteristic-correlation curve in combination with the historical correlation of the corresponding test data;
[0026] Analyzing the changes in the characteristic-correlation curve and calculating the characteristic-correlation dependence coefficient;
[0027] The feature association evaluation index of the test data is obtained by calculating the weighted average of all acquired feature-related dependency coefficients;
[0028] Obtain the feature difference value of the test data at the current moment and the previous moment, and then calculate the evaluation correlation between the current test data and the industrial Internet identifier by combining the historical correlation and feature association evaluation index at the previous moment;
[0029] Compare the evaluation correlation with the test analysis result output after the current test data is input into the initial analysis model to obtain a first correlation difference;
[0030] If the first correlation difference is greater than the set difference threshold, determining that the model test result of the current test data is negative;
[0031] Otherwise, it is determined that the model test result of the current test data is positive;
[0032] At this point, the calculation formula for the model performance index is as follows:
[0033] ; Where P represents the model performance index of the initial analysis model; It is expressed as the number of data with positive model test results; Represented as the total number of test data; The corresponding first correlation difference of the data with positive model test results is expressed as; It is expressed as the actual correlation between the test data with positive model test results and the industrial Internet logo; It is expressed as the weight value of the influence of the accuracy of model correlation analysis on the performance index of the calculated model; It is expressed as the number of data with negative model test results; It is represented by the corresponding first correlation difference of the data with negative model test results; It is expressed as the actual correlation between the test data with negative model test results and the industrial Internet logo; It is expressed as the weight value of the impact of the inaccuracy of model correlation analysis on the calculated model performance index.
[0034] Preferably, analyzing the change of the feature-correlation curve and calculating the feature-correlation dependence coefficient comprises:
[0035] Calculating the area under the first curve between each historical target characteristic value and the previous historical target characteristic value according to the characteristic-correlation curve;
[0036] Obtain a first area ratio of the area under the first curve to the area under the previous first curve;
[0037] Performing variance calculation on all acquired first area ratios to obtain a first dependence index;
[0038] If the first dependency index is not greater than the preset dependency threshold, then the first dependency index is used as the feature-correlation dependency coefficient of the correlation between the current target feature and the test data and the industrial Internet identifier;
[0039] If the first dependency index is greater than a preset dependency threshold, dividing the characteristic-correlation curve into a plurality of segmented characteristic curves according to the curve trend;
[0040] Combining and analyzing the characteristic-correlation curve and the segmented characteristic curve to calculate a second dependency index;
[0041] The calculation formula of the second dependency index is as follows:
[0042] ; In the formula, R represents the second dependence index; It is expressed as the corresponding eigenvalue of the i-th peak in the characteristic-correlation curve; It is expressed as the corresponding characteristic value of the i-th peak and valley in the characteristic-correlation curve; It is expressed as the time interval between the corresponding characteristic value of the i-th peak and the corresponding characteristic value of the i-th valley in the characteristic-correlation curve; It is represented as the i-th peak in the characteristic-correlation curve; It is represented as the i-th peak-valley in the characteristic-correlation curve; It is expressed as the average value of the ratio of the difference between the characteristic value difference and the correlation difference corresponding to the adjacent peaks and valleys that have not been calculated in the current characteristic-correlation curve; It is expressed as the average curve slope of the segmented characteristic curve with an upward curve trend; It is expressed as the average curve slope of the segmented characteristic curve with a downward curve trend; It is expressed as the influence factor of the curve trend change; It is expressed as the number of segmented characteristic curves with rising curve trend; It is expressed as the number of segmented characteristic curves with a downward trend;
[0043] The first dependency index and the second dependency index are combined to calculate the feature-related dependency coefficient.
[0044] Preferably, optimizing the model parameters of the initial analysis model includes:
[0045] Set the initial population size, maximum number of iterations, and upper and lower bounds of parameters;
[0046] Introduce the reverse refraction learning mechanism to initialize the population;
[0047] Set the fitness function and calculate the fitness value of the individual;
[0048] Based on the obtained fitness values, the optimal individual position is determined and saved;
[0049] An adaptive t-distribution mutation operator is introduced to perform t-distribution mutation on the position of the optimal individual based on the number of iterations;
[0050] Determine whether the maximum number of iterations has been reached or the convergence condition has been met. If so, output the global optimal solution and bring the optimized parameters into the initial analysis model to obtain the associated analysis model.
[0051] Preferably, obtaining the final search results based on the initial search results and parsing them to obtain the identified content includes:
[0052] Sort the target iconic character segments from large to small according to the initial search results, obtain the ordered iconic fields and output them as the final search results;
[0053] The final search result is analyzed using the Industrial Internet identification resolution platform to obtain the corresponding identification content.
[0054] Compared with the prior art, the beneficial effects of this application are as follows:
[0055] By training a neural network to build an association analysis model, we analyze the correlation between the target iconic character segment selected from the ordinary character string containing the industrial Internet logo of building materials equipment and the industrial Internet logo, and parse the analysis results to obtain the content of the logo, which effectively improves the accuracy of identifying the building materials equipment logo from the ordinary label string and the speed of extracting the logo information.
[0056] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.
[0057] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0058] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:
[0059] Figure 1 This is a flowchart of a method for extracting industrial Internet building materials equipment identification based on a neural network in an embodiment of the present invention. DETAILED DESCRIPTION
[0060] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.
[0061] The embodiment of the present invention provides a method for extracting the identification of building materials equipment in the industrial Internet based on a neural network, such as Figure 1 As shown, including:
[0062] Step 1: Obtain a common string containing the building materials equipment industrial Internet logo, and select several target iconic character segments from it;
[0063] Step 2: Analyze the correlation between the target iconic character segment and the industrial Internet logo using the association analysis model obtained by training the neural network system based on the effective logo data in the industrial Internet platform to obtain the initial search results;
[0064] Step 3: Obtain and parse the final search results based on the initial search results to obtain the identified content.
[0065] In this embodiment, the logos created by different building materials companies in the building materials equipment Industrial Internet logo have different corporate prefixes, while the logos created by the same building materials company have the same corporate prefix; the ordinary character string containing the building materials equipment Industrial Internet logo is diverse, and may be a character string that carries the Industrial Internet logo into a certain page of the WeChat mini program, or it may be a character string that has identification information and also contains other information; the target iconic character segment is a character segment selected from the ordinary character string containing the building materials equipment Industrial Internet logo based on the classification of the building materials equipment Industrial Internet logo information.
[0066] In this embodiment, the neural network system refers to a recurrent neural network and a convolutional neural network connected in series, wherein the recurrent neural network is used to process the temporal characteristics of the building materials and equipment identification information; the convolutional neural network is used to process the location characteristics of the building materials and equipment identification information; the association analysis model is used to analyze the correlation between the target iconic character segment and the industrial Internet identification, wherein the industrial Internet identification is created by the industrial Internet identification resolution secondary node platform.
[0067] In this embodiment, the initial search result refers to the correlation between the target iconic character segment input into the association analysis model and the output and the industrial Internet logo; the final search result refers to the ordered iconic field obtained by sorting the target iconic character segment from large to small according to the initial search result.
[0068] The beneficial effect of the above technical solution is: by training a neural network to build an association analysis model to analyze the correlation between the target iconic character segment selected from the ordinary character string containing the industrial Internet logo of building materials equipment and the industrial Internet logo, and parsing the analysis results to obtain the content of the logo, it effectively improves the accuracy of identifying the building materials equipment logo from the ordinary label character string and the speed of extracting the logo information.
[0069] The embodiment of the present invention provides a method for extracting an industrial Internet building material equipment identifier based on a neural network, which obtains a common character string containing an industrial Internet building material equipment identifier and selects a number of target iconic character segments therefrom, including:
[0070] Get a common string containing the industrial Internet identifier of building materials equipment;
[0071] According to the classification of the building materials equipment industrial Internet identification information, several character segments are screened from the obtained common character strings and output as target iconic character segments.
[0072] In this embodiment, the ordinary character string containing the Industrial Internet identification of building materials equipment is diverse. It may be a character string that carries the Industrial Internet identification to enter a certain page of the WeChat mini program, or it may be a character string that has identification information and also contains other information; the classification of the Industrial Internet identification information of building materials equipment includes name, parameters, color, Mac address, and header, etc.; the target iconic character segment is a character segment selected from the ordinary character string containing the Industrial Internet identification of building materials equipment based on the classification of the Industrial Internet identification information of building materials equipment.
[0073] The beneficial effect of the above technical solution is: by obtaining a common character string containing the building materials equipment industrial Internet identification, the target iconic character segment is screened out based on the classification of the building materials equipment industrial Internet identification information, thereby providing data support for subsequent identification content analysis.
[0074] The embodiment of the present invention provides a method for extracting industrial Internet building materials and equipment identification based on a neural network, using an association analysis model obtained by training a neural network system based on effective identification data in an industrial Internet platform, to analyze the correlation between the target iconic character segment and the industrial Internet building materials and equipment identification, and obtain several initial search results, including:
[0075] Performing data preprocessing on valid identification data obtained from the industrial Internet platform to obtain a first data set;
[0076] Divide the first data set according to a preset division ratio to obtain training data and test data;
[0077] After adding an additive attention mechanism and a CBAM attention mechanism to the recurrent neural network and the convolutional neural network respectively, the training data is used to train the neural network to obtain an initial analysis model;
[0078] Inputting the test data into the initial analysis model to obtain test analysis results;
[0079] Based on the test analysis results, determining a model performance index of the initial analysis model;
[0080] When the model performance index is not greater than a set performance threshold, optimizing the model parameters of the initial analysis model to obtain a correlation analysis model;
[0081] Otherwise, outputting the initial analysis model as a correlation analysis model;
[0082] The target iconic character segment is input into the association analysis model to obtain the correlation between the target iconic character segment and the industrial Internet building materials equipment identification, and output it as the initial search result.
[0083] In this embodiment, valid identification data refers to the valid character segment corresponding to the valid identification information; the first data set is a data set obtained by preprocessing the valid identification data obtained from the industrial Internet platform, wherein the data preprocessing includes removing duplicate or invalid data in the original data, performing data cleaning, data standardization, data normalization, and then performing data integration and clustering to create labels for the data; the preset division ratio is set in advance.
[0084] In this embodiment, the role of the additive attention mechanism is to effectively summarize important information in the time series and improve the learning ability of the model; the CBAM attention mechanism is an attention mechanism module that combines space and channels, and is used to enhance the representation ability of the convolutional neural network; the initial analysis model is a model obtained by training the series-connected recurrent neural network and convolutional neural network using training data; the test analysis result is the correlation between the test data and the industrial Internet logo output after inputting the initial analysis model; the model performance index is used to express the initial analysis model's ability to analyze the correlation between the iconic character segment and the industrial Internet logo.
[0085] In this embodiment, the performance threshold is set in advance; the initial search result refers to the target landmark character segment input into the association analysis model, and the output of the association between the target landmark character segment and the industrial Internet building materials equipment identification.
[0086] The beneficial effect of the above technical solution is: by using the association analysis model obtained by training the neural network system based on the effective identification data in the industrial Internet platform, the correlation between the target landmark character segment and the industrial Internet building materials equipment identification can be accurately obtained, which is conducive to ensuring that the subsequent analysis of the identification content is accurate and effective.
[0087] The embodiment of the present invention provides a method for extracting building material equipment identification of industrial Internet based on a neural network, and based on the test analysis result, determines the model performance index of the initial analysis model, including:
[0088] Extract historical target feature data of current test data within a preset historical time period and the historical correlation with the industrial Internet identifier from the identifier extraction database;
[0089] Determine the historical target characteristic value based on the historical target characteristic data, and construct the characteristic-correlation curve in combination with the historical correlation of the corresponding test data;
[0090] Analyzing the changes in the characteristic-correlation curve and calculating the characteristic-correlation dependence coefficient;
[0091] The feature association evaluation index of the test data is obtained by calculating the weighted average of all acquired feature-related dependency coefficients;
[0092] Obtain the feature difference value of the test data at the current moment and the previous moment, and then calculate the evaluation correlation between the current test data and the industrial Internet identifier by combining the historical correlation and feature association evaluation index at the previous moment;
[0093] Compare the evaluation correlation with the test analysis result output after the current test data is input into the initial analysis model to obtain a first correlation difference;
[0094] If the first correlation difference is greater than the set difference threshold, determining that the model test result of the current test data is negative;
[0095] Otherwise, it is determined that the model test result of the current test data is positive;
[0096] At this point, the calculation formula for the model performance index is as follows:
[0097] ; Where P represents the model performance index of the initial analysis model; It is expressed as the number of data with positive model test results; Represented as the total number of test data; The corresponding first correlation difference of the data with positive model test results is expressed as; It is expressed as the actual correlation between the test data with positive model test results and the industrial Internet logo; It is expressed as the weight value of the influence of the accuracy of model correlation analysis on the performance index of the calculated model; It is expressed as the number of data with negative model test results; It is represented by the corresponding first correlation difference of the data with negative model test results; It is expressed as the actual correlation between the test data with negative model test results and the industrial Internet logo; It is expressed as the weight value of the impact of the inaccuracy of model correlation analysis on the calculated model performance index.
[0098] In this embodiment, the identifier extraction database is composed of valid identifier data, historical features, and historical relevance within the industrial Internet platform, where the historical relevance is used to express the historical relevance between the valid character segment and the industrial Internet identifier.
[0099] In this embodiment, the historical preset time period is set in advance; the historical target feature data refers to the historical feature data of the target feature of the test data, wherein the target feature refers to time series feature, state feature, location feature, etc.
[0100] In this embodiment, the feature-correlation curve refers to a curve constructed by using the historical target feature values determined by the historical target feature data of the test data and combining them with the corresponding historical correlation; the feature-correlation dependency coefficient is used to express the influence of the target feature on the correlation between the test data and the industrial Internet identifier; the feature association evaluation index is used to express the total influence of all target features of the test data on the correlation with the industrial Internet identifier.
[0101] In this embodiment, the evaluation relevance refers to the correlation between the test data at the current moment and the industrial Internet identifier; the first correlation difference refers to the difference between the evaluation correlation and the correlation between the test analysis results output after the current test data is input into the initial analysis model; the setting difference threshold is set in advance.
[0102] The beneficial effect of the above technical solution is: by evaluating the correlation between the test data and the industrial Internet identifier at the current moment, and comparing it with the test analysis results to calculate the model performance index, an effective evaluation of the correlation analysis capability of the initial analysis model is achieved.
[0103] The embodiment of the present invention provides an industrial Internet building material equipment identification extraction method based on a neural network, analyzing the change of the feature-correlation curve, and calculating the feature-correlation dependency coefficient, including:
[0104] Calculating the area under the first curve between each historical target characteristic value and the previous historical target characteristic value according to the characteristic-correlation curve;
[0105] Obtain a first area ratio of the area under the first curve to the area under the previous first curve;
[0106] Performing variance calculation on all acquired first area ratios to obtain a first dependence index;
[0107] If the first dependency index is not greater than the preset dependency threshold, then the first dependency index is used as the feature-correlation dependency coefficient of the correlation between the current target feature and the test data and the industrial Internet identifier;
[0108] If the first dependency index is greater than a preset dependency threshold, dividing the characteristic-correlation curve into a plurality of segmented characteristic curves according to the curve trend;
[0109] Combining and analyzing the characteristic-correlation curve and the segmented characteristic curve to calculate a second dependency index;
[0110] The calculation formula of the second dependency index is as follows:
[0111] ; In the formula, R represents the second dependence index; It is expressed as the corresponding eigenvalue of the i-th peak in the characteristic-correlation curve; It is expressed as the corresponding characteristic value of the i-th peak and valley in the characteristic-correlation curve; It is expressed as the time interval between the corresponding characteristic value of the i-th peak and the corresponding characteristic value of the i-th valley in the characteristic-correlation curve; It is represented as the i-th peak in the characteristic-correlation curve; It is represented as the i-th peak-valley in the characteristic-correlation curve; It is expressed as the average value of the ratio of the difference between the characteristic value difference and the correlation difference corresponding to the adjacent peaks and valleys that have not been calculated in the current characteristic-correlation curve; It is expressed as the average curve slope of the segmented characteristic curve with an upward curve trend; It is expressed as the average curve slope of the segmented characteristic curve with a downward curve trend; It is expressed as the influence factor of the curve trend change; It is expressed as the number of segmented characteristic curves with rising curve trend; It is expressed as the number of segmented characteristic curves with a downward trend;
[0112] The first dependency index and the second dependency index are combined to calculate the feature-related dependency coefficient.
[0113] In this embodiment, the characteristic-correlation curve refers to a curve constructed by using the historical target characteristic values determined by the historical target characteristic data of the test data and combining them with the corresponding historical correlation; the first area ratio refers to the area ratio of the area under the first curve to the area under the previous first curve, wherein the area under the first curve refers to the area under the curve between the current historical target characteristic value and the previous historical target characteristic value; the first dependence index is obtained by performing variance calculation on all the first area ratios obtained.
[0114] In this embodiment, the preset dependency threshold is set in advance; the feature-correlation dependency coefficient is used to express the degree of influence of the target feature on the correlation between the test data and the industrial Internet identifier; the curve trends include rising, falling and stable; the second dependency index is obtained by combining the feature-correlation curve and the segmented feature curve for analysis, wherein the segmented feature curve is a segmented curve obtained by dividing the feature-correlation curve based on the curve trend.
[0115] The beneficial effect of the above technical solution is: by analyzing the changes in the feature-correlation curve and calculating the feature-correlation dependency coefficient, the degree of influence of the characteristics of the test data on the calculation of the correlation between the test data and the industrial Internet identification can be determined, thereby helping to improve the accuracy of identifying building materials equipment identification from ordinary label strings.
[0116] The embodiment of the present invention provides a method for extracting building material equipment identification of industrial Internet based on a neural network, which optimizes the model parameters of the initial analysis model, including:
[0117] Set the initial population size, maximum number of iterations, and upper and lower bounds of parameters;
[0118] Introduce the reverse refraction learning mechanism to initialize the population;
[0119] Set the fitness function and calculate the fitness value of the individual;
[0120] Based on the obtained fitness values, the optimal individual position is determined and saved;
[0121] An adaptive t-distribution mutation operator is introduced to perform t-distribution mutation on the position of the optimal individual based on the number of iterations;
[0122] Determine whether the maximum number of iterations has been reached or the convergence condition has been met. If so, output the global optimal solution and bring the optimized parameters into the initial analysis model to obtain the associated analysis model.
[0123] In this embodiment, the reverse refraction learning mechanism is used to expand the population search range, which is beneficial to improving the optimization effect of the algorithm; the adaptive t-distribution mutation operator is used to avoid falling into local optimality in the later stage of iteration and improve the parameter optimization effect; the association analysis model is a model obtained by optimizing the model parameters of the initial analysis model to obtain the optimized parameters, which is used to analyze the correlation between the target landmark character segment and the industrial Internet logo, wherein the model parameters include the learning rate of the LSTM-CNN neural network, the offset variable, the weighting coefficients of multiple levels, etc.
[0124] The beneficial effect of the above technical solution is: by optimizing the parameters of the initial analysis model to obtain the optimized parameters, an association analysis model is obtained to ensure the accuracy of the correlation between the target iconic character segment and the industrial Internet, which is conducive to improving the accuracy of subsequent identification content analysis.
[0125] The embodiment of the present invention provides a method for extracting the identification of building materials equipment in the industrial Internet based on a neural network, obtaining and parsing the final search results according to the initial search results to obtain the content of the identification, including:
[0126] Sort the target iconic character segments from large to small according to the initial search results, obtain the ordered iconic fields and output them as the final search results;
[0127] The final search result is analyzed using the Industrial Internet identification resolution platform to obtain the corresponding identification content.
[0128] In this embodiment, the target iconic character segment is a character segment selected from the common character segments containing the industrial Internet logo of building materials equipment; the initial search result refers to the relevance of the target iconic character segment input into the association analysis model and the industrial Internet logo output; the final search result refers to the ordered iconic field obtained by sorting the target iconic character segment from large to small according to the initial search results.
[0129] In this embodiment, the industrial Internet identification resolution platform refers to a platform that connects all production factors in the industry, such as equipment, machines, materials, etc., by establishing a unified identification system. By connecting fragmented data and applications through the resolution system, it is possible to grasp information such as data sources, flow processes, and usage trends, thereby realizing the full life cycle management of the entire industrialized production.
[0130] The beneficial effect of the above technical solution is: by utilizing the industrial Internet identifier resolution platform to sort the target iconic character segments from large to small according to the initial search results, and parsing the obtained ordered iconic fields, the identifier content can be accurately and effectively obtained, thereby realizing the extraction of the industrial Internet identifier from an ordinary character string.
[0131] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.
Claims
1. A neural network-based method for extracting identification of industrial Internet building materials equipment. It is characterized in that include: Step 1: Obtain a common string containing the building materials equipment industrial Internet logo, and select several target iconic character segments from it; Step 2: Analyze the correlation between the target iconic character segment and the industrial Internet logo using the association analysis model obtained by training the neural network system based on the effective logo data in the industrial Internet platform to obtain the initial search results; Step 3: Obtain and parse the final search results based on the initial search results to obtain the marked content; Among them, the correlation analysis model obtained by training the neural network system based on the effective identification data in the industrial Internet platform is used to analyze the correlation between the target iconic character segment and the industrial Internet logo, and the initial search results are obtained, including: Performing data preprocessing on valid identification data obtained from the industrial Internet platform to obtain a first data set; Divide the first data set according to a preset division ratio to obtain training data and test data; After adding an additive attention mechanism and a CBAM attention mechanism to the recurrent neural network and the convolutional neural network respectively, the training data is used to train the neural network to obtain an initial analysis model; Inputting the test data into the initial analysis model to obtain test analysis results; Based on the test analysis results, determining a model performance index of the initial analysis model; When the model performance index is not greater than a set performance threshold, optimizing the model parameters of the initial analysis model to obtain a correlation analysis model; Otherwise, outputting the initial analysis model as a correlation analysis model; Input the target iconic character segment into the association analysis model, obtain the association between the target iconic character segment and the industrial Internet building materials equipment identifier, and output it as the initial search result; Wherein, based on the test analysis result, determining the model performance index of the initial analysis model includes: Extract historical target feature data of current test data within a preset historical time period and the historical correlation with the industrial Internet identifier from the identifier extraction database; Determine the historical target characteristic value based on the historical target characteristic data, and construct the characteristic-correlation curve in combination with the historical correlation of the corresponding test data; Analyzing the changes in the characteristic-correlation curve and calculating the characteristic-correlation dependence coefficient; The feature association evaluation index of the test data is obtained by calculating the weighted average of all acquired feature-related dependency coefficients; Obtain the feature difference value of the test data at the current moment and the previous moment, and then calculate the evaluation correlation between the current test data and the industrial Internet identifier by combining the historical correlation and feature association evaluation index at the previous moment; Compare the evaluation correlation with the test analysis result output after the current test data is input into the initial analysis model to obtain a first correlation difference; If the first correlation difference is greater than the set difference threshold, determining that the model test result of the current test data is negative; Otherwise, it is determined that the model test result of the current test data is positive; At this point, the calculation formula for the model performance index is as follows: ; Where P represents the model performance index of the initial analysis model; It is expressed as the number of data with positive model test results; Represented as the total number of test data; The corresponding first correlation difference of the data with positive model test results is expressed as; It is expressed as the actual correlation between the test data with positive model test results and the industrial Internet logo; It is expressed as the weight value of the influence of the accuracy of model correlation analysis on the performance index of the calculated model; It is expressed as the number of data with negative model test results; It is represented by the corresponding first correlation difference of the data with negative model test results; It is expressed as the actual correlation between the test data with negative model test results and the industrial Internet logo; It is expressed as the weight value of the impact of the inaccuracy of model correlation analysis on the calculated model performance index.
2. According to the neural network-based industrial Internet building materials equipment identification extraction method according to claim 1, It is characterized in that Get a common string containing the building materials equipment industrial Internet logo, and select several target iconic character segments from it, including: Get a common string containing the industrial Internet identifier of building materials equipment; According to the classification of the building materials equipment industrial Internet identification information, several character segments are screened from the obtained common character strings and output as target iconic character segments.
3. According to the neural network-based industrial Internet building materials equipment identification extraction method according to claim 2, It is characterized in that The industrial Internet identifier is created by the industrial Internet identifier resolution secondary node platform.
4. According to the neural network-based industrial Internet building materials equipment identification extraction method according to claim 1, It is characterized in that Analyzing the change of the feature-correlation curve and calculating the feature-correlation dependence coefficient include: Calculating the area under the first curve between each historical target characteristic value and the previous historical target characteristic value according to the characteristic-correlation curve; Obtain a first area ratio of the area under the first curve to the area under the previous first curve; Performing variance calculation on all acquired first area ratios to obtain a first dependence index; If the first dependency index is not greater than the preset dependency threshold, then the first dependency index is used as the feature-correlation dependency coefficient of the correlation between the current target feature and the test data and the industrial Internet identifier; If the first dependency index is greater than a preset dependency threshold, dividing the characteristic-correlation curve into a plurality of segmented characteristic curves according to the curve trend; Combining and analyzing the characteristic-correlation curve and the segmented characteristic curve to calculate a second dependency index; The calculation formula of the second dependency index is as follows: ; In the formula, R represents the second dependence index; It is expressed as the corresponding eigenvalue of the i-th peak in the characteristic-correlation curve; It is expressed as the corresponding characteristic value of the i-th peak and valley in the characteristic-correlation curve; It is expressed as the time interval between the corresponding characteristic value of the i-th peak and the corresponding characteristic value of the i-th valley in the characteristic-correlation curve; It is represented as the i-th peak in the characteristic-correlation curve; It is represented as the i-th peak-valley in the characteristic-correlation curve; It is expressed as the average value of the ratio of the difference between the characteristic value difference and the correlation difference corresponding to the adjacent peaks and valleys that have not been calculated in the current characteristic-correlation curve; It is expressed as the average curve slope of the segmented characteristic curve with an upward curve trend; It is expressed as the average curve slope of the piecewise characteristic curve with the curve trend being downward; It is expressed as the influence factor of the curve trend change; It is expressed as the number of segmented characteristic curves with rising curve trend; It is expressed as the number of segmented characteristic curves with a downward trend; The first dependency index and the second dependency index are combined to calculate the feature-related dependency coefficient.
5. According to the neural network-based industrial Internet building materials equipment identification extraction method according to claim 1, It is characterized in that Optimizing the model parameters of the initial analysis model includes: Set the initial population size, maximum number of iterations, and upper and lower bounds of parameters; Introduce the reverse refraction learning mechanism to initialize the population; Set the fitness function and calculate the fitness value of the individual; Based on the obtained fitness values, the optimal individual position is determined and saved; An adaptive t-distribution mutation operator is introduced to perform t-distribution mutation on the position of the optimal individual based on the number of iterations; Determine whether the maximum number of iterations has been reached or the convergence condition has been met. If so, output the global optimal solution and bring the optimized parameters into the initial analysis model to obtain the associated analysis model.
6. According to the neural network-based industrial Internet building materials equipment identification extraction method according to claim 1, It is characterized in that The final search results are obtained and parsed according to the initial search results to obtain the marked content, including: Sort the target iconic character segments from large to small according to the initial search results, obtain the ordered iconic fields and output them as the final search results; The final search result is analyzed using the Industrial Internet identification resolution platform to obtain the corresponding identification content.
