Complex structure test data analysis method and system

Abstract

The invention relates to a complex structure test data analysis method and system. The method comprises the following steps: S1, constructing a metadata template library through a dynamic table; s2, constructing a data analysis model according to the characteristics of the test data, wherein the data analysis model is used for performing data analysis according to the metadata in the metadata template library; s3, training the data analysis model; s4, optimizing and adjusting the data analysis model according to the performance of the data analysis model and the characteristics of the test data; step S5, deploying the trained data analysis model in a data platform; and S6, performing automatic analysis on the complex structure test data through the data analysis model. According to the method, automatic analysis of complex structure test data is carried out by constructing the metadata template library and the data analysis model, unified data acquisition, storage and management of the data can be realized, data of different systems are integrated into a central database, and sharing and interaction of the data are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of test data analysis, and in particular to a method and system for analyzing test data of a complex structure. Background Art

[0002] In view of the current situation of scattered, isolated, dispersed and low utilization of data in the current enterprise informatization construction, it is proposed to build a dual-middle-end architecture based on the data middle-end and the business middle-end. The data middle-end helps to solve the problems of business data storage, circulation and application. With the popularization and application of the global acquisition system in various units, how to analyze the test data in the test data package has become an urgent problem to be solved. Although the manual application of envelope analysis, normal analysis, and tolerance analysis of product success data in the process of model product development has achieved certain results, due to the complex structure test In order to accurately obtain structural data, it is usually necessary to set up multiple measurement points, and manual analysis has obvious limitations in terms of large data volume, unclear indicators, and untimely analysis. Therefore, how to provide a system for realizing tasks such as automatic data cleaning, processing, and calculation, and provide better data analysis functions for aerospace product test data, has become an urgent problem to be solved. Summary of the invention

[0003] In order to solve the technical problems existing in the above-mentioned prior art, the purpose of the present invention is to provide a complex structure test data analysis method to solve the problem of complex structure test data analysis.

[0004] To achieve the above-mentioned object of the invention, the present invention provides a complex structure test data analysis method, characterized in that it comprises the following steps:

[0005] Step S1: construct a metadata template library through a dynamic table, converge the data source, and complete the consistency and integrity check of the metadata template library;

[0006] Step S2: constructing a data analysis model according to the characteristics of the test data, wherein the data analysis model is used to perform data analysis according to the metadata in the metadata template library;

[0007] Step S3, training the data analysis model;

[0008] Step S4: optimizing and adjusting the data analysis model according to the performance of the data analysis model and the characteristics of the test data;

[0009] Step S5: deploy the trained data analysis model on the data platform;

[0010] Step S6: Automatically analyzing the complex structure test data using the data analysis model.

[0011] According to a technical solution of the present invention, in step S3, it specifically includes:

[0012] Step S31, obtaining training data and preprocessing the training data; the training data is historical data of complex structure test data and manually annotated analysis results thereof; the preprocessing includes data cleaning, data noise reduction, data integration, data deduplication, data reduction and data normalization, and the data cleaning includes missing value processing, outlier detection and processing of the training data;

[0013] Step S32: Select a data analysis model according to business needs, taking into account data size and complexity, aluminum bar performance requirements, characteristic nonlinear relationship requirements, and interpretability requirements;

[0014] Step S33: training the data analysis model using the preprocessed training data;

[0015] Step S34: performing a performance evaluation on the trained data analysis model according to evaluation indicators, wherein the evaluation indicators include accuracy, precision, recall rate and F1 score.

[0016] According to a technical solution of the present invention, in step S4, according to the poor performance in the evaluation result of the data analysis model, one or several optimization methods are selected to optimize the model; the model optimization methods include:

[0017] Feature selection: according to the evaluation results of the model evaluation, the trained data analysis model performs poorly, the training set is optimized to obtain the optimal feature subset, and the poorly performing data analysis model is retrained;

[0018] Hyperparameter adjustment: according to the evaluation results of the model evaluation, the trained data analysis model performs poorly, and the hyperparameter combination of the poorly performing data analysis model is optimized, and the data analysis model is retrained; the hyperparameters include learning rate, number of iterations, and regularization strength;

[0019] Model structure adjustment, including adjustment of the number of layers, number of neurons, and activation function of the data analysis model;

[0020] Ensemble learning: After the data analysis model is trained, the stability and accuracy of the model are improved through ensemble methods;

[0021] Deep learning: Use a specific task or data set to further train the data analysis model and adjust the weights of the data analysis model to adapt to the specific task or data set.

[0022] According to a technical solution of the present invention, in step S5, it specifically includes:

[0023] Step S51, constructing the architecture of the complex structure test data analysis system based on the data analysis model;

[0024] Step S52, the system architecture determined in step S51, and coding of the complex structure test data analysis system;

[0025] Step S53, deploying the complex structure test data analysis system on the data platform, and performing a comprehensive test on the complex structure test data analysis system;

[0026] Step S54, designing data security and privacy protection for the complex structure test data analysis system;

[0027] Step S55: provide technical support and maintenance for the complex structure test data analysis system.

[0028] According to a technical solution of the present invention, in the step S51, it specifically includes:

[0029] Step S511: Build a data middle platform service, which is used to collect and calculate the test data, unify the data to form standard data and store it;

[0030] Step S512: Construct a test data analysis process around the identification and analysis of key characteristics of the product;

[0031] Step S513: Build a visualization function for the data center for data display.

[0032] According to a technical solution of the present invention, in step S6, it specifically includes:

[0033] Step S61, constructing a data analysis template according to the complex structure test analysis requirements and the metadata template library;

[0034] Step S62: review the data analysis template according to actual business information, template compilation status, and whether there is metadata available for analysis;

[0035] Step S63, generating a collection task and performing data collection of complex structure test data according to the data analysis template after review;

[0036] Step S64: perform data analysis on the collected data using the data analysis model and generate data analysis results.

[0037] According to a technical solution of the present invention, in step S62, the review of the data analysis template includes:

[0038] Completeness review: Ensure that all necessary data fields are included in the data analysis template;

[0039] Consistency review: ensure that the fields in the data analysis template are consistent with the fields in the data source;

[0040] Accuracy review: Ensure that the logic and calculation formulas in the data analysis template are correct.

[0041] According to a technical solution of the present invention, the metadata is stored in a table form, including the project name, data type, data attributes, whether automatic interpretation is required, and whether data application is required; wherein the project name includes the detection point name, detection method, judgment criteria, detection results, conclusions and remarks; and the data attributes are required values ​​or measured values.

[0042] According to one aspect of the present invention, a complex structure test data analysis system is provided, which is generated based on the above-mentioned complex structure test data analysis method and is used for automatically analyzing complex structure test data.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] The present invention proposes a complex structure test data analysis method and system, which summarizes the means of identifying, controlling and analyzing product quality technical risks from the practice of China's aerospace engineering. Through test data analysis, this method reflects the transformation of China's aerospace quality management from qualitative to quantitative, not only improving the accuracy of test data analysis from the original 85% to 95%, but also shortening the time of data processing and analysis from the original 20 hours to 5 hours, and improving the objectivity and reliability of product quality assessment, the yield rate from the original 70% of manual judgment to 90%, and at the same time enhancing the effect of aerospace model risk identification and control, from the previous manual analysis to system automatic warning.

[0045] Enterprises can use the data analysis model trained by the method provided by the present invention to perform data analysis, which can better understand business processes and optimize decision-making, thereby improving the efficiency and competitiveness of the enterprise. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0047] Figure 1Schematically showing a flow chart of complex structure test data analysis provided in one embodiment of the present invention;

[0048] Figure 2 Schematically showing a flow chart of automatic analysis model training according to one embodiment of the present invention;

[0049] Figure 3 Schematically showing a flow chart of data analysis according to one embodiment of the present invention;

[0050] Figure 4 A schematic diagram schematically showing metadata according to an embodiment of the present invention;

[0051] Figure 5 A schematic diagram schematically shows analysis results according to an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The description of the embodiments of this specification should be combined with the corresponding drawings, which should be considered as part of the complete specification. In the drawings, the shape or thickness of the embodiments may be enlarged and indicated for simplification or convenience. Furthermore, the parts of each structure in the drawings will be described separately. It is worth noting that the elements not shown in the drawings or not described in words are in a form known to ordinary technicians in the relevant technical field.

[0053] The description of the embodiments herein and any reference to directions and orientations are only for the convenience of description and are not to be construed as any limitation on the scope of protection of the present invention. The following description of the preferred embodiments may involve combinations of features, which may exist independently or in combination, and the present invention is not particularly limited to the preferred embodiments. The scope of the present invention is defined by the claims.

[0054] like Figure 1 As shown, the present invention provides a complex structure test data analysis method, comprising the following steps:

[0055] Step S1: construct a metadata template library through a dynamic table, converge the data source, and complete the consistency and integrity check of the metadata template library;

[0056] The metadata template library needs to consider the following rules during construction:

[0057] 1. Data source: Ensure that the data comes from real business scenarios, and the metadata to be analyzed should be of numerical type.

[0058] 2. Data integrity: Check whether the data is complete, and the metadata needs to have rules for handling missing values.

[0059] 3. Data consistency: Ensure consistency in data format and units.

[0060] 4. Data association: association rules between different data items.

[0061] The consine algorithm is used to implement string matching to ensure that the field names, data types, etc. of the metadata in the metadata template library can accurately match the corresponding items in the actual data.

[0062] The specific implementation method is:

[0063] Step S11, data vectorization: converting the metadata in the metadata template library into a vector form that can perform mathematical operations.

[0064] This usually involves tokenization and vectorization of the metadata text data. For example, you can use CountVectorizer to convert the metadata text data into a vector, where each dimension represents a word and the value represents the frequency of the word in the document.

[0065] Step S12, calculating cosine similarity: using the cosine_similarity function to calculate the cosine similarity between vectors.

[0066] This function returns a similarity matrix, where each element represents the similarity between two vectors.

[0067] Step S13, setting a similarity threshold: determining a similarity threshold for judging whether two metadata are similar enough to be considered as duplicates.

[0068] This threshold can be set according to the needs of specific applications.

[0069] Step S14, determining duplicate metadata: traversing the similarity matrix, for each pair of metadata, if their similarity values ​​exceed a preset threshold, it is considered that the metadata already exists, and the data storage operation is not performed.

[0070] like Figure 4 As shown, the metadata in the metadata template library is stored in a table format, including the project name, data type, data attributes, whether automatic interpretation is required, and whether data application is required; the project name includes the detection point name, detection method, judgment criteria, detection results, conclusions and remarks; the data attributes are required values ​​or measured values.

[0071] Step S2: construct a data analysis model according to the characteristics of the test data, and the data analysis model is used to perform data analysis according to the metadata;

[0072] Step S3, training the data analysis model;

[0073] like Figure 2 As shown, in step S3, it specifically includes:

[0074] Step S31, obtaining training data and preprocessing the training data; the training data is the historical data of complex structure test data and the analysis results of its manual annotation; the preprocessing includes data cleaning, data noise reduction, data integration, data deduplication, data reduction and data normalization, and data cleaning includes missing value processing, outlier detection and processing of training data;

[0075] Step S32: Select a data analysis model based on business needs, taking into account data size and complexity, aluminum bar performance requirements, characteristic nonlinear relationship requirements, and interpretability requirements.

[0076] The following example illustrates the idea of ​​selecting a data analysis model:

[0077] 1. Data size and complexity: For small-scale data sets, simple models such as linear regression or decision trees may be more effective. For large-scale data sets, we may need to choose models that can handle high-dimensional data and are computationally efficient. For example, in the manufacturing industry, predictive maintenance is an important application scenario, which involves a large amount of equipment sensor data. In this case, supervised learning models such as random forests or support vector machines that can handle large-scale data can be selected for fault prediction.

[0078] 2. Robustness requirement: If the model needs to be highly robust to outliers or noise, support vector machine (SVM) can be selected because it improves the generalization ability of the model by maximizing the margin of the classification boundary. For example, in the manufacturing industry, equipment failure prediction needs to be robust to abnormal equipment status, so SVM can be a suitable choice. If the robustness requirement is not high, linear regression or polynomial regression can be considered.

[0079] 3. Nonlinear relationship of features: When there is a complex nonlinear relationship between the dependent variable and the independent variable of the data features, a polynomial regression model or a deep learning model (such as a convolutional neural network CNN or a recurrent neural network RNN) can be selected. For example, in quality control in the manufacturing industry, using CNN for image recognition and defect detection can capture complex patterns on the surface of the product. If the relationship is roughly linear, linear regression or SVR can be selected.

[0080] 4. Explanatory requirements: If the business needs require the model to be highly interpretable, that is, to be able to provide intuitive explanations to better understand how the model makes predictions, then decision trees or rule-based classification models may be more appropriate because they can provide clear decision paths and rules. In the manufacturing industry, for example, in the product design stage, it may be necessary to understand how different design parameters affect product performance. At this time, using a decision tree as a proxy model can provide an intuitive analysis of the impact of design parameters. If interpretability is not a key requirement, SVM or polynomial regression can be considered.

[0081] Step S33: training the data analysis model using the preprocessed training data;

[0082] Step S34: Perform performance evaluation on the trained data analysis model. The specific steps of performance evaluation are as follows:

[0083] 1. Prepare the evaluation dataset: Divide the dataset into training set, validation set, and test set. Make sure that the data in the test set is not used in the training process to ensure the fairness of the evaluation results.

[0084] 2. Select evaluation metrics: Based on the model’s objectives and business needs, select appropriate evaluation metrics such as accuracy, precision, recall, and F1 score.

[0085] 3. Implement cross-validation: Use methods such as K-fold cross-validation to ensure that the model performs consistently on different subsets.

[0086] 4. Model evaluation: Run the model on the validation set, record the results of each evaluation metric, analyze the performance of the model on different categories, and check whether there is any class imbalance problem.

[0087] 5. Result analysis: Conduct in-depth analysis of the evaluation results to identify the strengths and weaknesses of the model performance and visualize the results of the model predictions through tools such as confusion matrix.

[0088] 6. Error analysis: Analyze the model’s incorrect predictions to understand the types and causes of errors.

[0089] When training a data analysis model, focus on the following parameters:

[0090] 1. Scalars:

[0091] (1) Loss: Monitor the changes in loss during training and validation to evaluate the learning effect of the model.

[0092] (2) Accuracy: Monitor the changes in the model’s accuracy to understand the accuracy of the model’s predictions.

[0093] (3) Learning Rate: Changes in the learning rate are critical to model training and need to be monitored to ensure that it is adjusted as expected.

[0094] 2. Performance evaluation parameters:

[0095] (1) Accuracy: Accuracy is the ratio of correctly classified samples to the total number of samples. It is an intuitive indicator of the overall performance of the model, that is, the model’s ability to make correct predictions.

[0096] (2) Precision: Precision refers to the proportion of positive cases predicted to be positive and actually positive. It can measure the accuracy of the model in predicting positive cases, that is, the ability to reduce false positive cases.

[0097] (3) Recall: Recall refers to the ratio of samples predicted as positive to samples actually positive. It can measure the ability of the model to detect all positive samples, that is, the ability to reduce false negative examples.

[0098] (4) F1 Score: The F1 score is the harmonic mean of precision and recall, that is, 2*(precision*recall) / (precision+recall). The F1 score is a useful metric when a balance is needed between precision and recall.

[0099] Step S4: Optimizing and adjusting the data analysis model according to the performance of the data analysis model and the characteristics of the test data;

[0100] According to the poor performance of the evaluation results of the data analysis model, one or more of the following optimization methods are selected to optimize the model:

[0101] 1) Feature selection: According to the evaluation results of the model evaluation, the trained data analysis model performs poorly. The training set is optimized to obtain the optimal feature subset, and the data analysis model with poor performance is retrained. When performing feature optimization to obtain the optimal feature subset, it can be achieved through filtering, packaging or embedding methods.

[0102] Filter methods evaluate the relationship between each feature and the target variable based on statistical tests. Commonly used statistical measures include correlation coefficient, chi-square test, and mutual information. Filter methods are simple and fast, but may not capture the interaction between features.

[0103] Wrapper Methods construct different feature subsets and use machine learning models to evaluate the performance of these subsets to select the best feature subset. This method can capture the combined effect between features, but the computational cost is high.

[0104] Embedded methods automatically perform feature selection during model training. This method is often used in conjunction with certain machine learning models to assess the importance of features through weights or importance scores during model learning.

[0105] Feature selection specifically includes the following steps:

[0106] a. Feature importance evaluation: Use tree models and other methods to evaluate the importance of features;

[0107] b. Recursive feature elimination: Select a subset of features by recursively removing the least important features;

[0108] c. Model-based feature selection: Use the model’s coefficients or feature importance for feature selection.

[0109] If the model performs poorly, it may be due to improper feature selection. You can optimize the model through feature selection. The goal of feature selection is to find the optimal feature subset that can achieve the best model performance on the training set.

[0110] 2) Hyperparameter adjustment: According to the evaluation results of the model evaluation, the trained data analysis model performs poorly, and the hyperparameter combination of the poorly performing data analysis model is optimized, and the data analysis model is retrained; in machine learning, hyperparameters are a set of parameters set before training begins, which have an important impact on the performance of the model. Common hyperparameters include learning rate, number of iterations, regularization strength, etc. The optimal hyperparameter combination can be found through methods such as grid search, random search, or Bayesian optimization.

[0111] Hyperparameter adjustment includes adjustment of learning rate, batch size, and number of iterations:

[0112] Learning rate: If model training is unstable or converges slowly, adjusting the learning rate may help improve model performance;

[0113] Batch size: Changing the batch size can affect the convergence speed and stability of the model;

[0114] Iterations: Increase or decrease the number of iterations to ensure that the model has enough training time or to avoid overfitting.

[0115] 3) Adjustment of model structure, including adjustment of the number of layers and neurons:

[0116] Number of layers and neurons: Increase or decrease the number of layers and neurons in the model to adapt to the complexity of the data.

[0117] Activation function: Try different activation functions to improve the nonlinear fitting ability of the model.

[0118] The data analysis model pruning, quantization and other technologies can be used to reduce model parameters and computational complexity, thereby improving the reasoning speed of the data analysis model.

[0119] 4) Regularization technology: Use L1, L2 regularization and other techniques to prevent overfitting of data analysis models.

[0120] 5) Ensemble learning: After the data analysis model is trained, the stability and accuracy of the model can be further improved through ensemble methods (such as bagging, boosting, stacking, etc.);

[0121] The specific steps include:

[0122] Select base model: Depending on the nature of the problem (regression, classification or clustering) and the characteristics of the data, a set of different base models are selected. These models can be decision trees, support vector machines, random forests, gradient boosting machines, neural networks, etc.

[0123] Train base models: Train these base models separately. Each model should be trained on the same training dataset to ensure that their predictions can be reasonably compared and combined.

[0124] Selection of integration strategy: To improve stability, choose to reduce the variance of the model through Bagging (such as random forest); to improve the accuracy of the model, choose Boosting (such as AdaBoost, XGBoost) to focus on samples that were misclassified by the previous model; to improve the prediction effect of the model, use Stacking to train a new model to learn how to best combine the predictions of the basic models.

[0125] Model fusion: Use the selected ensemble strategy to fuse the predictions of the base models. Model fusion can be performed by weighted averaging, voting mechanism, or training a new meta-learner.

[0126] Parameter tuning and optimization: Tune and optimize the parameters of the ensemble model to further improve performance. This can be achieved by adjusting the number of base models, the complexity of each base model, and / or the parameters of the ensemble strategy.

[0127] Performance evaluation: Evaluate the performance of the ensemble model on the validation set to ensure that the model is not overfitting or underfitting, and compare it with the performance of a single base model.

[0128] 6) Deep learning: Further training (fine-tuning) using a specific task or data set involves adjusting the weights of the model to adapt to the new task or data.

[0129] The specific steps include:

[0130] Pre-trained model selection: Choose a pre-trained deep learning model as a starting point, which can leverage features learned on large-scale datasets.

[0131] Transfer of learning results: Apply the pre-trained model to a specific task through transfer learning, freeze the weights of some layers, and add or replace some layers on the top layer to adapt to the new task.

[0132] Training and fine-tuning: The deep learning model is trained using the training dataset. During the training process, the weights of the model can be adjusted through fine-tuning to improve the performance of the model on a specific task.

[0133] Performance evaluation and tuning: Evaluate the performance of the model on the validation set and perform tuning as needed. This may include adjusting the network architecture, adding data augmentation, changing optimizer parameters, etc.

[0134] 7) Early termination: During the training of the data analysis model, if the performance of the data analysis model on the validation set no longer improves, the training can be stopped early.

[0135] Step S5: deploy the trained data analysis model on the data platform;

[0136] In step S5, it specifically includes:

[0137] Step S51, constructing the architecture of the complex structure test data analysis system based on the data analysis model;

[0138] Wherein, step S51 specifically includes:

[0139] Step S511: Build a data middle platform service, which is used to collect and calculate test data, unify the data to form standard data and store it;

[0140] The data middle platform can collect, calculate, store and process data technology and massive data, while unifying standards and calibers. By unifying the data into standard data and then storing it, it forms big data assets and provides customers with efficient services.

[0141] Step S512: Construct a test data analysis process around the identification and analysis of key characteristics of the product;

[0142] The experimental data analysis process includes:

[0143] A) The user configures the template that the user needs to use according to the test requirements and metadata template library;

[0144] B) Create a new data table and use the data center service to synchronize business data to the newly created data table.

[0145] C) Use the data cleaning function in the data center service to clean, deduplicate, and delete data with problematic parameters in the data table;

[0146] D) Use the data analysis model to analyze and calculate the cleaned data, and store the calculation structure in a temporary intermediate table to facilitate users to obtain data.

[0147] Step S513: Build a visualization function for the data center for data display.

[0148] Through the visualization of the data center, it is convenient to display data, which makes it easier for users to analyze and apply data, and can realize functions such as data set access and analysis results management.

[0149] Step S52: The system architecture determined in step S51 is used to write the code for the complex structure test data analysis system;

[0150] Step S53: deploy the complex structure test data analysis system on the data platform, and conduct a comprehensive test on the complex structure test data analysis system;

[0151] System Testing and Deployment: The system is fully tested to ensure its stability and reliability, and then it can be deployed to the actual operating environment for actual application and testing.

[0152] Step S54, designing data security and privacy protection for the complex structure test data analysis system;

[0153] The system needs to be designed for data security and privacy protection to ensure the security and privacy of test data. This may involve data encryption, access control, security auditing and other aspects.

[0154] Step S55: Provide technical support and maintenance for the complex structure test data analysis system.

[0155] Through technical support and maintenance, we can provide continuous technical support to ensure the normal operation and use effect of the system.

[0156] Step S6: Automatically analyze the complex structure test data through the data analysis model.

[0157] like Figure 3 As shown, in step S6, it specifically includes:

[0158] Step S61, constructing a data analysis template according to complex structure test analysis requirements and metadata template library;

[0159] Step S62: Review the data analysis template based on actual business information, template compilation status, and whether there is metadata available for analysis. The review includes:

[0160] 1. Completeness of the template: Make sure all necessary data fields are included.

[0161] 2. Template consistency: Ensure that the fields in the template are consistent with the fields in the data source.

[0162] 3. Accuracy of the template: Ensure that the logic and calculation formulas in the template are correct.

[0163] Step S63: Generate a collection task and collect data of complex structure test data after reviewing the data analysis template;

[0164] Step S64: perform data analysis on the collected data through the data analysis model and generate data analysis results.

[0165] The generated data analysis results can be Figure 5 shown.

[0166] In addition, it should be noted that the present invention can be provided as a method, an apparatus or a computer program product. Therefore, the embodiments of the present invention can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the embodiments of the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code.

[0167] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0168] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable terminal device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0169] It should also be noted that, in this article, the terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0170] Finally, it should be noted that the above is a preferred embodiment of the present invention. It should be pointed out that although the preferred embodiment of the present invention has been described, for those skilled in the art, once the basic creative concept of the present invention is known, several improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications should also be regarded as the protection scope of the present invention. Therefore, the attached claims are intended to be interpreted as including the preferred embodiment and all changes and modifications that fall within the scope of the embodiments of the present invention.

Claims

1. A complex structure test data analysis method, characterized in that: The following steps are involved: Step S1: construct a metadata template library through a dynamic table, converge the data source, and complete the consistency and integrity check of the metadata template library; Step S2: constructing a data analysis model according to the characteristics of the test data, wherein the data analysis model is used to perform data analysis according to the metadata in the metadata template library; Step S3, training the data analysis model; Step S4: optimizing and adjusting the data analysis model according to the performance of the data analysis model and the characteristics of the test data; Step S5: deploy the trained data analysis model on the data platform; Step S6: Automatically analyzing the complex structure test data using the data analysis model.

2. The complex structure test data analysis method according to claim 1, characterized in that: In the step S3, it specifically includes: Step S31, obtaining training data and preprocessing the training data; the training data is historical data of complex structure test data and manually annotated analysis results thereof; the preprocessing includes data cleaning, data noise reduction, data integration, data deduplication, data reduction and data normalization, and the data cleaning includes missing value processing, outlier detection and processing of the training data; Step S32: Select a data analysis model according to business needs, taking into account data size and complexity, aluminum bar performance requirements, characteristic nonlinear relationship requirements, and interpretability requirements; Step S33: training the data analysis model using the preprocessed training data; Step S34: Perform performance evaluation on the trained data analysis model according to evaluation indicators, where the evaluation indicators include accuracy, precision, recall rate and F1 score.

3. The complex structure test data analysis method according to claim 1, characterized in that: In step S4, according to the poor performance of the evaluation result of the data analysis model, one or more optimization methods are selected to optimize the model; the model optimization methods include: Feature selection: according to the evaluation results of the model evaluation, the trained data analysis model performs poorly, the training set is optimized to obtain the optimal feature subset, and the poorly performing data analysis model is retrained; Hyperparameter adjustment: according to the evaluation results of the model evaluation, the trained data analysis model performs poorly, and the hyperparameter combination of the poorly performing data analysis model is optimized, and the data analysis model is retrained; the hyperparameters include learning rate, number of iterations, and regularization strength; Model structure adjustment, including adjustment of the number of layers, number of neurons, and activation function of the data analysis model; Ensemble learning: After the data analysis model is trained, the stability and accuracy of the model are improved through ensemble methods; Deep learning: Use a specific task or data set to further train the data analysis model and adjust the weights of the data analysis model to adapt to the specific task or data set.

4. The complex structure test data analysis method according to claim 1, characterized in that: In the step S5, it specifically includes: Step S51, constructing the architecture of the complex structure test data analysis system based on the data analysis model; Step S52, the system architecture determined in step S51, and coding of the complex structure test data analysis system; Step S53, deploying the complex structure test data analysis system on the data platform, and performing a comprehensive test on the complex structure test data analysis system; Step S54, designing data security and privacy protection for the complex structure test data analysis system; Step S55: provide technical support and maintenance for the complex structure test data analysis system.

5. The complex structure test data analysis method according to claim 4, characterized in that: In the step S51, it specifically includes: Step S511: Build a data middle platform service, which is used to collect and calculate the test data, unify the data to form standard data and store it; Step S512: Construct a test data analysis process around the identification and analysis of key characteristics of the product; Step S513: Build a visualization function for the data center for data display.

6. The complex structure test data analysis method according to claim 5, characterized in that: In the step S6, it specifically includes: Step S61, constructing a data analysis template according to the complex structure test analysis requirements and the metadata template library; Step S62: review the data analysis template according to actual business information, template compilation status, and whether there is metadata available for analysis; Step S63, generating a collection task and performing data collection of complex structure test data according to the data analysis template after review; Step S64: perform data analysis on the collected data using the data analysis model and generate data analysis results.

7. The complex structure test data analysis method according to claim 6, characterized in that: In step S62, the review of the data analysis template includes: Completeness review: Ensure that all necessary data fields are included in the data analysis template; Consistency review: ensure that the fields in the data analysis template are consistent with the fields in the data source; Accuracy review: Ensure that the logic and calculation formulas in the data analysis template are correct.

8. The complex structure test data analysis method according to claim 1, characterized in that: The metadata is stored in a table format, including the project name, data type, data attributes, whether automatic interpretation is required, and whether data application is required; wherein the project name includes the detection point name, detection method, judgment criteria, detection results, conclusions, and remarks; and the data attributes are required values ​​or measured values.

9. A complex structure test data analysis system, characterized in that: The method is generated based on the complex structure test data analysis method as described in any one of claims 1 to 8 and is used for automatically analyzing complex structure test data.

Images