Product quality detection method and device based on data item grading and machine learning, equipment and medium

Abstract

The invention discloses a product quality detection method and device based on data item grading and machine learning, equipment and a medium, and relates to the technical field of data processing, and the method comprises the steps: obtaining original data related to product quality; dividing data items in the original data into a plurality of grades based on manual intervention; the levels include reserved decision terms, larger impact terms, and optionally reserved smaller impact terms; obtaining a grading result, and carrying out quality evaluation and cleaning and repairing processing on the data of the reserved decision items, the data of the relatively large influence items and the data of the relatively small influence items; carrying out feature extraction and feature selection on the processed data to obtain feature data having significant influence on product quality, and training a machine learning model for product quality evaluation to obtain a product quality evaluation model; and inputting data of a to-be-detected product into the product quality evaluation model, and outputting a product quality evaluation result. According to the invention, the efficiency and precision of product quality detection based on machine learning are improved.

Description

Product quality inspection methods, devices, equipment, and media based on data item classification and machine learning. Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a product quality inspection method, apparatus, equipment, and medium based on data item classification and machine learning. Background Technology

[0002] In the field of machine learning-based product quality inspection, mainstream technologies typically follow a standardized data processing flow: "Data Acquisition → Data Cleaning → Feature Engineering → Model Training." Specifically, firstly, multi-source data related to the entire product manufacturing process is collected through various sensors deployed on the production line or related business information systems (such as MES, ERP, etc.). This includes raw material attributes, process parameters, equipment operating status, and finished product appearance and performance inspection data. Subsequently, a series of automated data cleaning and feature engineering algorithms are used to denoise and reduce the dimensionality of the raw data. Finally, the processed data is input into machine learning models based on neural networks for training to complete tasks such as defect identification or quality grade classification. During this process, to automate data cleaning and feature extraction, algorithms automatically filter noise and extract key features. For example, in image detection, Gaussian filtering and median filtering are commonly used for image denoising, and autoencoders or gradient boosting trees are used for feature dimensionality reduction. In signal processing, Fourier transform and wavelet transform are used to remove noise frequency components, and recursive feature elimination is used to select key signal features.

[0003] However, the above model essentially places the intervention point of domain knowledge at the back end of the process, while relying on the autonomous decision-making of the algorithm in the early critical stages of data preprocessing.

[0004] The lag in domain knowledge intervention is reflected in the fact that human intervention only exists in the data labeling or model parameter tuning stages, while data cleaning and feature engineering rely on automatic processing by algorithms. When dealing with complex industrial data, there is often a lack of deep understanding of the business background and physical meaning, which makes it impossible to effectively distinguish between core data that has a decisive impact on product quality and irrelevant interference data. A large amount of interference information still remains in the cleaned data. As a result, subsequent machine learning models have to face high-dimensional, low signal-to-noise ratio input data, which not only increases the computational burden and overfitting risk of the model, but also makes the model have to learn effective features from noisy data with great difficulty, reducing learning efficiency and accuracy.

[0005] While some automated feature engineering methods can reduce data dimensionality, they require complex calculations on the entire raw data set. This computational time increases exponentially, especially when dealing with massive industrial-scale inspection data. Furthermore, high-dimensional data leads to an excessive number of parameters in machine learning models, exacerbating the risk of overfitting and necessitating reliance on high-performance hardware such as GPUs, thus driving up deployment costs.

[0006] The above factors make it difficult to further improve the overall performance of the quality inspection system. Summary of the Invention

[0007] This invention provides a product quality inspection method, apparatus, equipment, and medium based on data item classification and machine learning, which solves the problem of how to improve efficiency and accuracy in product quality inspection based on machine learning.

[0008] To achieve the above objectives, this application adopts the following technical solution: Firstly, it provides a product quality inspection method based on data item grading and machine learning, comprising: acquiring raw data related to product quality; dividing the data items in the raw data into multiple levels based on manual intervention; the levels include retained decision items, items with greater impact, and optional items with lesser impact; acquiring grading results, performing quality assessment, cleaning, and repair processing on the data of the retained decision items, items with greater impact, and items with lesser impact; performing feature extraction and feature selection on the processed data to obtain feature data that significantly affects product quality, and training a machine learning model for product quality evaluation to obtain a product quality evaluation model; inputting the data of the product to be tested into the product quality evaluation model, and outputting the product quality evaluation result.

[0009] Furthermore, the grading also includes insignificant items; the data of the less impactful items and / or insignificant items are configured to be optionally removed; the data of the less impactful items is also used to repair the data of the decision items and the more impactful items.

[0010] Furthermore, the quality assessment, cleaning, and repair of the retained decision items, major impact items, and minor impact items includes: performing the following quality assessments: data integrity assessment, data accuracy assessment, data redundancy assessment, and data noise assessment; removing abnormal data based on the results and classification results of the quality assessments; repairing missing or noisy data in the decision items and major impact items after removing abnormal data; and quantitatively and qualitatively verifying the quality of the repaired data.

[0011] Furthermore, based on the results of the quality assessment and the grading results, abnormal data is removed; wherein, for data with missing fields, the threshold for data removal is dynamically adjusted so that the field missing rate removal threshold used for the decision item is greater than the field missing rate removal threshold used for items with greater or lesser impact.

[0012] Secondly, a product quality inspection device based on data item grading and machine learning is provided, comprising: a raw data acquisition module for acquiring raw data related to product quality; a data item division and data processing module for dividing data items in the raw data into multiple levels based on manual intervention; the levels include retained decision items, items with greater impact, and optional items with lesser impact; acquiring grading results, performing quality assessment, cleaning, and repair processing on the data of the retained decision items, items with greater impact, and items with lesser impact; a model training module for performing feature extraction and feature selection on the processed data to obtain feature data that has a significant impact on product quality, and training a machine learning model for product quality evaluation to obtain a product quality evaluation model; and a product quality inspection module for inputting the data of the product to be inspected into the product quality evaluation model and outputting the product quality evaluation result.

[0013] Thirdly, an electronic device is provided, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the product quality inspection method based on data item grading and machine learning as described in the first aspect.

[0014] Fourthly, a readable storage medium is provided, on which a program or instructions are stored, which, when executed by a processor, implement the steps of the product quality inspection method based on data item grading and machine learning as described in the first aspect. Attached Figure Description

[0015] Figure 1 is a schematic flowchart of a product quality inspection method based on data item classification and machine learning provided in an embodiment of this application. Detailed Implementation

[0016] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the technical solutions in the embodiments of this application are clearly described. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art are within the scope of protection of this application.

[0017] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0018] The steps described in the specification and the flowcharts in the accompanying drawings of this invention are not necessarily strictly executed according to the step numbers; the execution order of the method steps can be changed. Furthermore, certain steps can be omitted, multiple steps can be combined into one step, and / or one step can be broken down into multiple steps.

[0019] Existing technologies lack a domain knowledge intervention mechanism before data cleaning, resulting in high data dimensionality, strong noise interference, and low computational efficiency, making it difficult to meet the needs of high-precision, real-time quality detection.

[0020] Based on this, this specification provides a product quality inspection method based on data item classification and machine learning. This method incorporates domain knowledge before data cleaning, aiming to proactively screen key data dimensions and eliminate noise sources through a domain knowledge intervention mechanism before data cleaning, thereby reducing data complexity from the source and improving the training efficiency and detection accuracy of the detection model. It also relates to a corresponding product quality inspection device based on data item classification and machine learning, a computer device, and a computer-readable storage medium, which will be described in detail below with reference to the accompanying drawings and preferred embodiments.

[0021] Please refer to Figure 1. This application embodiment provides a product quality inspection method based on data item classification and machine learning, as shown in Figure 1, including: Step S1: Obtain raw data related to product quality.

[0022] This step collects product quality-related data from multiple sources, including various sensors on the product production line, production management systems, and quality testing equipment. This data includes raw material parameters, production process parameters, equipment operating status data during production, and product appearance and performance testing data.

[0023] Step S2: Based on manual intervention, the data items in the original data are divided into multiple levels; the levels include retained decision items, items with greater impact, and optional items with lesser impact; obtain the classification results, and perform quality assessment, cleaning, and repair processing on the data of the retained decision items, items with greater impact, and items with lesser impact.

[0024] This step is a crucial part of the data quality management and analysis process. Its core purpose is to optimize data quality from the source by incorporating expert experience into the early stages of data preprocessing, thereby providing more efficient and cleaner input data for subsequent model training.

[0025] By professionally assessing and categorizing data items according to their impact, we can focus on core value and optimize analysis efficiency.

[0026] For example, the following explanation covers four aspects: grading standards, evaluation mechanisms, retention strategies, and flexible adjustments. Regarding grading standards, data items are divided into four levels based on their impact on product quality analysis, specifically defined as follows: Decisive Items: Core data items that play a decisive role in the conclusions of product quality analysis. Their absence or errors will directly lead to analysis failure or serious deviations in conclusions, such as core product performance indicators, safety and compliance parameters, and key user behavior data.

[0027] Significantly Influencing Items: Secondary core data items that have a significant impact on the product quality analysis conclusions. Bias in these items may cause the conclusions to deviate from expectations, but can be corrected by supplementing other data, such as: secondary function usage rate, user experience rating, and non-core process time.

[0028] Minor impact items: Marginal data items that have a limited impact on the product quality analysis conclusions. They are usually used to supplement details, and their absence alone has no substantial impact on the conclusions (e.g., low-frequency function operation records, log data from unconventional scenarios).

[0029] Irrelevant items: Data items that are largely irrelevant to the product quality analysis objectives. They contribute nothing to the conclusions and only increase data processing costs (e.g., redundant fields, meaningless parameters inherited from previous versions).

[0030] Regarding the evaluation mechanism, the grading process should be led by product domain experts and completed in conjunction with actual analysis objectives. Experts should make comprehensive assessments based on dimensions such as the correlation between data items and product quality / business objectives, the potential risks of missing or incorrect data, and the criticality of business scenarios.

[0031] Evaluation methods: Delphi method (multi-round anonymous scoring), expert seminars, etc. can be used to ensure the professionalism and consensus of the grading results.

[0032] Regarding retention principles and flexible adjustments, after the classification is completed, the retention scope needs to be dynamically adjusted according to the analysis needs and data dimensions: In general scenarios, only the decisive items and items with significant impact should be retained in principle, focusing on core value and reducing interference from redundant data. This is suitable for special analyses with clear objectives, such as the diagnosis of key quality issues.

[0033] Flexible expansion: If the data dimensions are too small (e.g., small sample size, insufficient indicator coverage), smaller influencing items can be added and retained to avoid biased analysis due to insufficient data dimensions. This is suitable for comprehensive evaluation scenarios that require multi-dimensional cross-validation.

[0034] Optional removal: Irrelevant items can be removed directly according to actual needs to reduce storage and computing costs; items with minor impact can be retained if analysis resources are sufficient, otherwise they can be removed as needed.

[0035] In some possible implementations, the categorization of levels may also include insignificant items; the data of the less impactful items and / or insignificant items may be optionally removed; the data of the less impactful items may also be used to repair the data of the decision items and the more impactful items.

[0036] Furthermore, the data that has been graded undergoes data processing. The core objective is to systematically clean and repair the data to eliminate anomalies, redundancies, and missing data, thereby improving data integrity and reliability and providing high-quality input for subsequent product quality analysis. The processing must balance "accurately removing invalid data" and "reasonably repairing high-value data." Specifically, the quality assessment, cleaning, and repair of the retained decision items, major impact items, and minor impact items includes: Step S21: Performing the following quality assessments: data integrity assessment, data accuracy assessment, data redundancy assessment, and data noise assessment.

[0037] First, a multi-dimensional quality assessment is conducted on the graded dataset to clarify the types and scope of anomalies that need to be addressed. This includes the following assessment dimensions: 1. Data integrity assessment: Statistical analysis of the missing rate (number of missing records / total number of records) of each data item (especially "decisive items" and "items with significant impact"), and marking fields (e.g., single field missing rate > 30%) or records (e.g., single record missing key fields > 50%).

[0038] Example: If 300 records of "Product Core Performance Indicator A" (determinant) are missing out of 1000 records, special attention should be paid to how such records are handled.

[0039] 2. Data accuracy assessment: Check whether the data conforms to business logic and format specifications: Logical errors: such as "product price = 0 yuan", "user age = -5 years old", "fault occurrence time is later than the reporting time", etc., which violate common sense; Format errors: such as date fields containing "2023-13-01", numeric fields containing text characters (such as "123abc"); Range errors: such as "user satisfaction rating" exceeding the preset range (such as 6 points appearing in the 0-5 score range).

[0040] 3. Data redundancy assessment identifies completely duplicate records (all field values ​​are the same) or duplicate records in key fields (such as duplicate "product ID + test batch" combinations) to avoid interference from duplicate data with analysis results.

[0041] 4. Data Noise Assessment: For numerical data (such as temperature and pressure values ​​collected by sensors), abnormal fluctuations (such as a sudden increase in temperature of 100°C in a single day followed by a return to normal) are identified using methods such as box plots and Z-scores, and records containing noise are marked.

[0042] Step S22: Based on the results of the quality assessment and the grading results, remove abnormal data; wherein, for data with missing fields, dynamically adjust the data removal threshold so that the field missing rate removal threshold used for the decision item is greater than the field missing rate removal threshold used for items with greater or lesser impact.

[0043] Based on the quality assessment results, data records that are "obviously erroneous, duplicated, or severely missing" are removed to prioritize ensuring the basic reliability of the dataset.

[0044] 1. Principles for handling obvious errors: For "hard errors" that violate logic, format, or scope, directly remove records that cannot be corrected; for errors that can be corrected by rules (such as standardizing date formats or converting text-based numerical values ​​to numbers), prioritize correction rather than removal.

[0045] Example: If a record has "System crash (hardware failure)" in the "Fault Type" field (which is logically contradictory, as hardware failure should not be classified as system crash), and the correct type cannot be inferred from the context, then the record should be removed; if the "Temperature" field contains "35℃" (a duplicate unit), it can be standardized to "35℃" using a regular expression, and there is no need to remove it.

[0046] 2. Duplicate record removal process: First, identify completely duplicate records (all field values ​​are the same) and keep only 1 record; then identify duplicate records with key fields (e.g., "Product ID + Test Time" uniquely identifies a test record; if this combination is duplicated, keep the latest or most complete record).

[0047] 3. Threshold setting for removing severely missing records: Dynamically adjusted based on data classification: For records with missing "decision item" fields: If the missing "decision item" field in a single record is >50%, remove it directly (because the core data cannot support the analysis); For records with missing "significantly influential" fields: If the combined missing "significantly influential" and "decision item" fields in a single record are >70%, remove it; For records with missing "minorly influential" fields: The threshold can be relaxed (e.g., retain records with a missing rate <80%) to avoid excessive removal leading to insufficient data dimensions.

[0048] Step S23: Repair missing or noisy data in the decision items and major impact items after removing abnormal data.

[0049] For the remaining "minor missing" or "noisy" data (especially "decisive items" and "significantly influential items") after data removal, technical means are used to repair them to avoid analytical bias caused by data simplification.

[0050] Step S24: Quantitatively and qualitatively verify the quality of the repaired data.

[0051] After the repair is completed, the data quality needs to be verified through multiple dimensions to ensure its integrity and reliability.

[0052] 1. Quantitatively validate the statistically processed dataset: whether the missing rate has decreased to an acceptable range (e.g., missing rate of key fields < 5%), and whether the proportion of outliers has decreased (e.g., outlier proportion in box plot < 1%); compare the data distribution before and after repair: check whether the distribution of numerical data is consistent and whether the frequency of categorical data is stable through histograms and kernel density estimation (KDE).

[0053] 2. Qualitative verification combined with business logic checks: Whether the repaired data conforms to the actual scenario (e.g., whether the repaired "product lifespan" is within a reasonable range); Domain expert confirmation: Invite product experts to review key fields (e.g., the repair results of "decision items") to ensure they conform to business understanding.

[0054] It is worth noting that the entire data processing workflow needs to be recorded during the cleaning and repair process described above: including the number of records removed and the reasons, the repair methods and parameters, and the changes in verification indicators, in order to form a "Data Processing Report" to ensure traceability.

[0055] Step S3: Perform feature extraction and feature selection on the processed data to obtain feature data that has a significant impact on product quality, and train a machine learning model for product quality evaluation to obtain a product quality evaluation model.

[0056] In this step, data mining and statistical analysis methods are used to extract features from the processed data. For numerical data, at least one of the following is extracted as a feature: mean, variance, maximum, minimum, median, skewness, and kurtosis. For time series data, at least one of the following is extracted as a feature: trend and periodicity. For categorical data, after it is encoded into numerical form, its frequency or proportion is calculated as a feature.

[0057] By using feature selection algorithms, such as information gain, mutual information, and ReliefF, features that have a significant impact on product quality are screened out, redundant and irrelevant features are removed, data dimensionality is reduced, and the training efficiency and accuracy of subsequent machine learning models are improved.

[0058] Based on the characteristics of the product quality evaluation task and the data features, select an appropriate machine learning model, such as support vector machine, random forest, neural network, etc.

[0059] The feature-engineered data is divided into training, validation, and test sets according to a certain ratio. The selected machine learning model is trained using the training set data. The model's hyperparameters are adjusted, and the model's performance is evaluated using the validation set data. Techniques such as cross-validation are used to improve the model's generalization ability and prevent overfitting.

[0060] Step S4: Input the data of the product to be tested into the product quality evaluation model and output the product quality evaluation result.

[0061] The trained machine learning model is deployed to the actual production environment to receive relevant data on newly produced products in real time. The data, after being processed by the data acquisition module and the feature engineering module, is input into the trained model, which outputs product quality evaluation results, such as product quality grade, whether it is qualified, and the type and probability of quality defects.

[0062] Based on the quality evaluation results, a quality report is generated. The report includes information such as the product's various quality indicators, evaluation results, comparative analysis with standard values, and changes in quality trends. Simultaneously, the quality evaluation results are fed back to the production department to enable timely adjustments and optimizations to the production process, thereby improving product quality.

[0063] This application innovatively categorizes data items into decisive items, items with significant impact, items with minor impact, and insignificant items. Dimensionality reduction and labeling are performed on the data before data cleaning, reducing the difficulty of data cleaning and improving the efficiency of data analysis while ensuring accuracy. Multiple data mining and statistical analysis methods are comprehensively used for feature extraction, and advanced feature selection algorithms are used to accurately screen key features, effectively reducing data dimensionality. This improves model training efficiency while ensuring the model accurately captures product quality characteristics. Based on different product quality evaluation tasks and data characteristics, machine learning models are rationally selected, and through scientific adjustment of hyperparameters and the use of cross-validation, the models achieve optimal performance, enabling in-depth mining and accurate evaluation of complex relationships in product quality.

[0064] Based on the above technical solution, the following advantages are achieved: expert experience is transformed into a classification of data items, enabling deep integration of domain knowledge; data defects are compensated for by labeling key dimensions, thereby enhancing data quality; by pre-classifying data and retaining relatively important data items, the original data dimensions can be reduced, significantly reducing subsequent computational complexity and greatly shortening model training time; by leveraging expert experience to actively eliminate known noise sources, the data signal-to-noise ratio is improved to a higher level, reducing interference in model learning in noisy environments and lowering the false negative and false positive rates.

[0065] Corresponding to the above-described product quality inspection method embodiment based on data item grading and machine learning, this application embodiment provides a product quality inspection device based on data item grading and machine learning, comprising: a raw data acquisition module for acquiring raw data related to product quality; a data item division and data processing module for dividing the data items in the raw data into multiple levels based on manual intervention; the levels include retained decision items, items with greater impact, and optional items with lesser impact; acquiring grading results, performing quality assessment, cleaning, and repair processing on the data of the retained decision items, items with greater impact, and items with lesser impact; a model training module for performing feature extraction and feature selection on the processed data to obtain feature data that has a significant impact on product quality, and training a machine learning model for product quality evaluation to obtain a product quality evaluation model; and a product quality inspection module for inputting the data of the product to be inspected into the product quality evaluation model and outputting the product quality evaluation result.

[0066] Furthermore, the grading also includes insignificant items; the data of the less impactful items and / or insignificant items are configured to be optionally removed; the data of the less impactful items is also used to repair the data of the decision items and the more impactful items.

[0067] Furthermore, the quality assessment, cleaning, and repair of the retained decision items, major impact items, and minor impact items includes: performing the following quality assessments: data integrity assessment, data accuracy assessment, data redundancy assessment, and data noise assessment; removing abnormal data based on the results and classification results of the quality assessments; repairing missing or noisy data in the decision items and major impact items after removing abnormal data; and quantitatively and qualitatively verifying the quality of the repaired data.

[0068] Furthermore, based on the results of the quality assessment and the grading results, abnormal data is removed; wherein, for data with missing fields, the threshold for data removal is dynamically adjusted so that the field missing rate removal threshold used for the decision item is greater than the field missing rate removal threshold used for items with greater or lesser impact.

[0069] The product quality inspection device based on data item classification and machine learning described above implements the steps and processes of the above-described product quality inspection method based on data item classification and machine learning, and achieves the same technical effect. To avoid repetition, these will not be repeated here.

[0070] Corresponding to the above embodiments of the product quality inspection method based on data item classification and machine learning, this application provides an electronic device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps and processes of the above embodiments of the product quality inspection method based on data item classification and machine learning, and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0071] Memory can be used to store software programs and various data. Memory can primarily include a first storage area for storing programs or instructions and a second storage area for storing data. The first storage area can store the operating system, application programs or instructions required for at least one function (such as sound playback, image playback, etc.). Furthermore, memory can include volatile memory or non-volatile memory, or both. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (Synchlink DRAM, SLDRAM), and direct memory bus RAM (DRRAM). The memory in the embodiments of this application includes, but is not limited to, these and any other suitable types of memory.

[0072] The processor may include one or more processing units; optionally, the processor integrates an application processor and a modem processor, wherein the application processor mainly handles operations related to the operating system, user interface, and applications, while the modem processor mainly handles wireless communication signals, such as a baseband processor. It is understood that the aforementioned modem processor may also not be integrated into the processor.

[0073] Corresponding to the above embodiments of the product quality inspection method based on data item classification and machine learning, this application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the steps and processes of the above embodiments of the product quality inspection method based on data item classification and machine learning, and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0074] The processor is the processor in the electronic device described in the above embodiments of this application. The readable storage medium includes a computer-readable storage medium, such as a computer read-only memory (ROM), random access memory (RAM), a magnetic disk, or an optical disk.

[0075] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0076] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0077] It is understood that the embodiments of this application have been described above in conjunction with the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. As those skilled in the art will know, various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, those skilled in the art, under the guidance or instruction of this application, can modify these features and embodiments to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, this invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of this invention.

Claims

1. A product quality inspection method based on data item grading and machine learning, characterized in that, include: Obtain raw data related to product quality; Based on manual intervention, the data items in the original data are divided into multiple levels; the levels include retained decision items, items with greater impact, and optional items with lesser impact; the classification results are obtained, and the data of the retained decision items, items with greater impact, and items with lesser impact are subjected to quality assessment, cleaning, and repair processing; feature extraction and feature selection are performed on the processed data to obtain feature data that has a significant impact on product quality, and a machine learning model for product quality evaluation is trained to obtain a product quality evaluation model; the data of the product to be tested is input into the product quality evaluation model, and the product quality evaluation result is output.

2. The product quality inspection method based on data item grading and machine learning according to claim 1, characterized in that, The grading also includes insignificant items; the data of the less impactful items and / or insignificant items are configured to be optionally removed; the data of the less impactful items is also used to repair the data of the decision items and the more impactful items.

3. The product quality inspection method based on data item grading and machine learning according to claim 1, characterized in that, The quality assessment, cleaning, and repair of the retained decision items, major impact items, and minor impact items includes: performing the following quality assessments: data integrity assessment, data accuracy assessment, data redundancy assessment, and data noise assessment; removing abnormal data based on the results and classification results of the quality assessments; repairing missing or noisy data in the decision items and major impact items after removing abnormal data; and quantitatively and qualitatively verifying the quality of the repaired data.

4. The product quality inspection method based on data item grading and machine learning according to claim 3, characterized in that, The process involves removing abnormal data based on the quality assessment and grading results; specifically, for data with missing fields, the threshold for data removal is dynamically adjusted so that the field missing rate removal threshold used for the decision item is greater than the field missing rate removal threshold used for items with greater or lesser impact.

5. A product quality inspection device based on data item grading and machine learning, characterized in that, include: The system includes: a raw data acquisition module for acquiring raw data related to product quality; a data item segmentation and processing module for segmenting data items in the raw data into multiple levels based on manual intervention; the levels include decision items to be retained, items with greater impact, and items with lesser impact that can be optionally retained; obtaining the grading results; performing quality assessment, cleaning, and repair processing on the data of the retained decision items, items with greater impact, and items with lesser impact; a model training module for extracting and selecting features from the processed data to obtain feature data that significantly affects product quality, and training a machine learning model for product quality evaluation to obtain a product quality evaluation model; and a product quality detection module for inputting the data of the product to be tested into the product quality evaluation model and outputting the product quality evaluation results.

6. The product quality inspection device based on data item classification and machine learning according to claim 5, characterized in that, The grading also includes insignificant items; the data of the less impactful items and / or insignificant items are configured to be optionally removed; the data of the less impactful items is also used to repair the data of the decision items and the more impactful items.

7. The product quality inspection device based on data item grading and machine learning according to claim 5, characterized in that, The quality assessment, cleaning, and repair of the retained decision items, major impact items, and minor impact items includes: performing the following quality assessments: data integrity assessment, data accuracy assessment, data redundancy assessment, and data noise assessment; removing abnormal data based on the results and classification results of the quality assessments; repairing missing or noisy data in the decision items and major impact items after removing abnormal data; and quantitatively and qualitatively verifying the quality of the repaired data.

8. The product quality inspection device based on data item classification and machine learning according to claim 7, characterized in that, The process involves removing abnormal data based on the quality assessment and grading results; specifically, for data with missing fields, the threshold for data removal is dynamically adjusted so that the field missing rate removal threshold used for the decision item is greater than the field missing rate removal threshold used for items with greater or lesser impact.

9. An electronic device, characterized in that, The electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the product quality inspection method based on data item grading and machine learning as described in any one of claims 1 to 4.

10. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps of the product quality inspection method based on data item grading and machine learning as described in any one of claims 1 to 4.

Images