Multi-level quality file management method based on big data

Abstract

The invention relates to a multi-level quality file management method based on big data, and the method comprises the steps: quantifying the relation between quality file information and a quality control result through building a random forest model, and recognizing key quality file information which has an important influence on the quality control result; a reference basis is provided for improving the quality file information for improving the quality control effect; the use frequency of the quality files is clustered by adopting a clustering analysis method, and the importance of the quality files is sorted according to a clustering result, so that classified storage of the importance of the quality files is realized, and the file management efficiency is improved. Through dual importance evaluation of key quality file information identification and importance grade division of the quality file, the problem of low management efficiency caused by incapability of evaluating the importance of the quality file in the prior art is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of quality file management and relates to a multi-level quality file management method based on big data. Background Art

[0002] Quality documents refer to documents that record and specify information related to product, service or process quality. They are important tools for communication and quality control within an organization or across organizations. These documents include quality manuals, procedure documents, operating instructions, product quality standards, testing technical specifications and standard methods, quality plans, quality records and test reports, etc. With the advancement of information technology, the electronic and digitalization of quality documents makes management more efficient and convenient. In the global competition, good quality document management helps enterprises meet the requirements of the international market for product quality, thereby enhancing their competitiveness.

[0003] However, existing technologies cannot evaluate the importance of quality documents, making it difficult for companies to identify which document information is critical to quality control during the management process. This can lead to a waste of resources, increased management costs due to unnecessary document review and maintenance, and important quality documents may be overlooked, affecting product quality and compliance. In addition, the lack of an effective evaluation mechanism can lead to wrong decisions and the inability to adjust and optimize quality management processes in real time, which may ultimately have a negative impact on customer satisfaction and corporate reputation. Summary of the invention

[0004] The present invention provides a multi-level quality file management method based on big data, which aims to identify key quality file information by establishing a relationship between quality file information and quality control result data, determine the importance ranking of quality files by combining the use frequency data such as file access records and modification records, and store and manage quality files in a hierarchical manner according to importance. This solves the problem that the existing technology cannot evaluate the importance of quality files, resulting in low management efficiency.

[0005] The purpose of the present invention can be achieved by the following technical solutions: The present application provides a multi-level quality document management method based on big data, comprising the following steps: S1. Collecting a large-scale sample of quality file data, wherein the quality file data includes quality file information and its corresponding quality control result data; S2. Establish a random forest model between quality document information and quality control results to identify key quality document information that significantly affects quality control results; S3. Real-time monitoring and acquisition of the usage frequency of quality documents; S4. Clustering the usage frequencies using a cluster analysis method, and dividing the importance levels of the quality files according to the clustering results; S5. Classify and store the quality documents according to the importance level, and generate a summary of key quality document information as a label for each quality document; S6. Model application: Obtain the quality files to be executed, extract key quality file information and input it into the random forest model, and output the predicted quality control results; S7. Adjust the quality documents to be executed based on the predicted quality control results.

[0006] Furthermore, the quality document information includes quality manuals, standard operating procedures, checklists, quality plans, quality records, non-conforming product handling procedures, change control records, risk management documents, training records, and customer feedback and complaint records.

[0007] Furthermore, the quality control results include non-conforming rate, defect density, inspection pass rate, customer complaint rate, return rate, on-time delivery rate, product life, compliance rate, customer satisfaction and statistical process control data.

[0008] Specifically, in step S2, the random forest model includes the following construction steps: S21, divide the data set into a training set and a test set; S22, randomly extracting several quality file information variables from the training set as explanatory variables and quality control results as response variables to generate a decision tree; S23, repeating the operations of steps S21-S22 until all quality file information variables are traversed to obtain a random forest containing large-scale decision trees; S24, calculate the prediction error of each decision tree in the random forest for the test set; S25. Sort the importance of quality document information variables according to the size of the prediction error; S26. Select the top 40% of quality document information variables in importance as key quality document information.

[0009] Furthermore, the usage frequency includes the number of people who visit the quality file and the number of times the quality file is modified.

[0010] Furthermore, in step S4, the cluster analysis method is configured as a K-means clustering method, comprising the following steps: S41, using normalization method to process the frequency data variable so that the data falls within the range of 0 to 1; S42, randomly select K data points from the data set as initial cluster centers; S43, for each data point of the quality file, calculating the Euclidean distance between it and each initial cluster center, and assigning each data point to the initial cluster center with the smallest Euclidean distance, to obtain an initial cluster; S44, respectively calculating the mean of all data points in the initial clusters, and using the mean as a new cluster center; S45, continuously repeating the process of allocating data points and updating cluster centers until a stop condition is met, and determining a final cluster; S46: Determine the importance level corresponding to each final cluster according to the numerical range of the final cluster.

[0011] Furthermore, in step S45, the stop condition is specifically that the cluster center no longer changes or reaches a preset update number.

[0012] Furthermore, in step S5, the quality files are classified and stored according to the importance level, including the following steps: S51, establishing storage structure: establishing different folder or directory storage structures according to importance levels; S52, automatic classification: automatically identify the frequency of use of quality documents, determine the importance level according to the frequency of use and store it in the corresponding storage structure; S53, Dynamic Monitoring: Set a specific time step to dynamically monitor the importance level through the usage frequency of the stored quality files. When the importance level changes, migrate the quality files to the storage structure corresponding to the importance level and keep the migration log.

[0013] Furthermore, in step S5, the generation of a summary of key quality file information as a label for each quality file includes the following steps: Assign numerical values ​​to categorical variables in key quality document information; Arrange the values ​​in the order of importance of the quality file information variables to obtain a numerical sequence summary; The numerical sequence summary is used as a label of the quality file, and annotations are generated for each numerical value of the label.

[0014] Furthermore, in step S6, the quality file to be executed is also compared with the key quality file information similarity of the classified and stored quality files to determine the importance level of the quality file to be executed and store it.

[0015] Furthermore, in step S7, adjusting the quality file to be executed according to the predicted quality control result includes the following steps: S71. When the predicted quality control result is lower than the preset value, check the key quality document information; S72. Adjust abnormal key quality document information according to quality control requirements; S73. Input the adjusted key quality document information into the random forest model to re-predict the quality control results; S74. Repeat steps S71-S73 until the predicted quality control result is not lower than the preset value and the quality file to be executed determines that it meets the execution conditions.

[0016] Beneficial effects of the present invention: (1) By establishing a random forest model to quantify the relationship between quality document information and quality control results, key quality document information that has a significant impact on quality control results is identified, providing a reference for improving quality document information to enhance quality control effectiveness; (2) Cluster analysis method is used to cluster the frequency of use of quality documents, and the importance of quality documents is ranked according to the clustering results, so as to achieve classified storage of the importance of quality documents and improve the efficiency of document management. The present invention solves the problem that the existing technology cannot evaluate the importance of quality documents, resulting in low management efficiency, by dual importance evaluation of key quality document information identification and importance level classification of quality documents. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0018] Figure 1 The present invention is a flowchart of a multi-level quality file management method based on big data.

[0019] Figure 2 The figure is a flowchart for constructing a random forest model in one embodiment of the present invention.

[0020] Figure 3 The present invention is a flowchart of classifying and storing quality files in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation mode, structure, characteristics and effects of the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments.

[0022] See also Figure 1-Figure 3 , the present application provides a multi-level quality file management method based on big data, comprising the following steps: S1. Collecting a large-scale sample of quality file data, wherein the quality file data includes quality file information and its corresponding quality control result data; In this embodiment, the quality file information related to quality control includes the following categories: 1) Quality Manual: Includes the company's quality policy, objectives and the framework of the quality management system, and clarifies the organization's quality standards and commitments.

[0023] 2) Standard Operating Procedure (SOP): Clarify specific operating processes and steps to ensure that operations meet quality requirements.

[0024] 3) Checklist: Used for product or process inspection to help record and confirm whether various quality requirements are met.

[0025] 4) Quality Plan: Describes the quality objectives, resource allocation, and implementation steps for a specific project or product.

[0026] 5) Quality records: including inspection and test results, audit records, training records, etc., providing traceability of quality activities.

[0027] 6) Procedure for handling non-conforming products: Specify the identification, recording, disposal and improvement measures for non-conforming products to ensure that problems are resolved in a timely manner.

[0028] 7) Change control records: Manage any changes to products, processes or documents, and ensure that changes are strictly approved and recorded.

[0029] 8) Risk management documentation: Assess quality-related risks and develop control measures to reduce potential quality issues.

[0030] 9) Training records: Record employees’ quality training to ensure they master relevant knowledge and skills.

[0031] 10) Customer feedback and complaint records: Collect customer feedback and complaints in order to analyze and improve the quality of products and services.

[0032] Quality control result indicators are key data used to evaluate the quality of products or services. They help companies monitor and improve quality management. Common quality control result indicators include: 1) Unqualified rate: reflects the proportion of unqualified products in the total products, used to evaluate the overall quality level.

[0033] 2) Defect density: The number of defects per unit of product or service, which helps identify the severity of quality problems.

[0034] 3) Inspection pass rate: The proportion of products that have passed the quality inspection, reflecting the effectiveness of the inspection process.

[0035] 4) Customer complaint rate: The ratio of the number of customer complaints about products or services to total sales volume is closely related to customer satisfaction.

[0036] 5) Repair rate: The proportion of products that need to be repaired due to quality problems to the total products, indicating problems in the production or service process.

[0037] 6) On-time delivery rate: The proportion of orders delivered on time, reflecting service quality and supply chain efficiency.

[0038] 7) Product life: The average service life of a product under normal conditions of use, reflecting the reliability of the product.

[0039] 8) Compliance rate: the proportion of products or services that comply with relevant industry standards and regulations to ensure that the products are legal and compliant.

[0040] 9) Customer Satisfaction: Customer satisfaction scores collected through surveys directly measure customers' experience with products and services.

[0041] 10) Statistical process control (SPC) data: Use tools such as control charts to monitor the stability and consistency of the production process.

[0042] Furthermore, the quality document information includes quality manuals, standard operating procedures, checklists, quality plans, quality records, non-conforming product handling procedures, change control records, risk management documents, training records, and customer feedback and complaint records.

[0043] Furthermore, the quality control results include non-conforming rate, defect density, inspection pass rate, customer complaint rate, return rate, on-time delivery rate, product life, compliance rate, customer satisfaction and statistical process control data.

[0044] S2. Establish a random forest model between quality document information and quality control results to identify key quality document information that significantly affects quality control results; The impact of quality document information on quality control results is significant and far-reaching. First, quality documents provide standard operating procedures and implementation procedures to ensure that all employees follow a unified process to perform quality control, thereby reducing variation and improving the consistency and reliability of results. These documents often contain clear key quality indicators (KPIs), set quantifiable goals for quality control, and make the evaluation of results clear and easy to track. Secondly, quality documents provide employees with necessary training and guidance to help them understand quality standards and control requirements, thereby improving their ability to perform quality control. The preventive measures and corrective actions recorded in quality documents can effectively reduce potential defects, directly reduce the production of non-conforming products, and thus improve the control effect. Furthermore, the existence of a good feedback mechanism and evaluation method enables enterprises to identify problems in the quality control process in a timely manner and make corresponding adjustments to optimize the management process. Finally, perfect quality documents ensure compliance, reduce quality problems caused by non-compliance with standards, and further improve the overall control effect. In short, quality document information ensures the efficiency and effectiveness of the quality control process through multi-faceted support, and ultimately helps to improve the overall quality of products and services.

[0045] However, in this embodiment, due to the excessive amount of quality file information involved, the degree of influence on the quality control results is different, and it is difficult for people to identify the key quality file information with a greater degree of influence. To this end, this embodiment uses the quality file information as the explanatory variable and the quality control result as the response variable, constructs a random forest model to rank the importance of the quality file information, and selects the quality file information with the highest importance as the key quality file information. Random forest is a classic machine learning algorithm that can be used for feature extraction and variable importance ranking. Random forest trains and predicts by constructing multiple decision trees, and each decision tree only uses a part of the training data and part of the features for modeling. By constructing multiple decision trees, a powerful classification or regression model is finally obtained, in which each decision tree has a weight contribution. Random forest can give the importance ranking of each input feature in the model, and the importance scores of these features are calculated based on their degree of influence on the model prediction results in the random forest. Usually, the feature importance score in the random forest is calculated based on the out-of-bag error.

[0046] Specifically, in step S2, the random forest model includes the following construction steps: Step S21, dividing the data set into a training set and a test set; When building a random forest model, you first need to divide the data set of quality document information and its corresponding quality control result data into a training set and a test set. The purpose of this is to ensure the effectiveness and generalization ability of the model. The training set will be used to build the model, and the model parameters will be continuously adjusted to better fit the patterns in the data. The test set will be used after the model is built to evaluate the performance of the model on data that has never been seen. A common division ratio is to use about 70% of the data as a training set and the remaining 30% as a test set. The random division method can ensure that the data distribution of the training set and the test set is as similar as possible, making the model evaluation more accurate.

[0047] Step S22, randomly extracting a number of quality file information variables from the training set as explanatory variables and quality control results as response variables to generate a decision tree; Next, a portion of quality file information variables are randomly selected from the training set. These variables will be used as explanatory variables to predict the response variable of quality control results. Random forest is an integrated model composed of multiple decision trees. Randomly extracting variables each time a decision tree is built can increase the diversity of the model and avoid overfitting the model to a specific combination of variables. Common decision tree construction algorithms such as the CART algorithm will continuously split the data set into multiple child nodes by selecting the best split point. This process continues until the stopping condition is met, such as the decision tree reaches the maximum depth or the number of samples in the node is less than a certain threshold.

[0048] Step S23, repeating the operations of steps S21-S22 until all quality file information variables are traversed to obtain a random forest containing a large-scale decision tree; In order to build a powerful random forest model, steps S21-S22 need to be repeated many times to generate a large number of decision trees. Each time the operation is repeated, different variables and samples are randomly selected from the training set, which ensures that each decision tree has a certain degree of independence and difference. By building large-scale decision trees, the model can better capture the complex relationships in the data and improve the accuracy and stability of the prediction. Usually the number of decision trees in a random forest can range from dozens to hundreds, and the specific number can be adjusted according to actual conditions.

[0049] Step S24: Calculate the prediction error of each decision tree in the random forest for the test set; When the random forest is built, use this model to predict the test set. For each decision tree in the random forest, calculate the error between its prediction results for the samples in the test set and the actual quality control results. Common error calculation methods include mean square error, mean absolute error, etc. These error indicators can reflect the predictive ability of each decision tree and provide a basis for the subsequent evaluation of variable importance.

[0050] Step S25: sorting the importance of quality file information variables according to the size of the prediction error; By calculating the prediction error of each decision tree for the test set, the importance of each quality file information variable in the model can be evaluated. The specific method is to calculate the average contribution of the variable in all decision trees in the random forest. For example, the number of splits and information gain of the variable in the decision tree construction process can be calculated, and then the results of all decision trees can be averaged. Based on these importance evaluation results, the quality file information variables can be ranked, and the variables with higher importance play a greater role in the prediction of quality control results.

[0051] Step S26: Select the top 40% quality document information variables in importance as key quality document information.

[0052] Finally, according to the importance of the variables, select the first certain proportion of quality document information variables as key quality document information. Here we select the first 40% of the variables, and this proportion can be adjusted according to the actual situation. Key quality document information has a greater impact on the quality control results. In the subsequent quality document management and analysis, we can focus on these variables to conduct more effective quality control and management.

[0053] S3. Real-time monitoring and acquisition of the usage frequency of quality documents; Furthermore, the usage frequency includes the number of people who visit the quality file and the number of times the quality file is modified.

[0054] In this embodiment, the frequency of use of a quality file reflects the importance of the file, the number of visitors reflects the degree of attention people pay to the quality file, and the number of modifications reveals the update frequency and actual use of the quality file, both of which can prove the importance of the quality file. Therefore, this embodiment performs cluster analysis on the frequency of use of quality files, and classifies quality files with the same frequency of use characteristics into the same importance level, thereby realizing the importance ranking of quality files.

[0055] S4. Clustering the usage frequencies using a cluster analysis method, and dividing the importance levels of the quality files according to the clustering results; Furthermore, in step S4, the cluster analysis method is configured as a K-means clustering method, comprising the following steps: Step S41: using a normalization method to process the usage frequency data variable so that the data falls within the range of 0 to 1; Before cluster analysis, it is important to normalize the frequency data variables. Normalization can unify data of different magnitudes to the same scale, avoiding some variables from dominating the clustering process due to their large values. Through normalization, the data is mapped to the range of 0 to 1, so that each variable has equal importance in subsequent calculations, improving the accuracy and reliability of clustering results.

[0056] Step S42: randomly selecting K data points from the data set as initial cluster centers; In order to start the K-means clustering algorithm, K data points need to be randomly selected from the data set as the initial cluster centers. The selection of these initial cluster centers has a certain degree of randomness, and different random selections may lead to different clustering results, but as the algorithm iterates, the clustering results will gradually stabilize. The value of K is predetermined, indicating that the data is to be divided into K clusters.

[0057] Step S43: for each data point of the quality file, calculate the Euclidean distance between it and each initial cluster center, and assign each data point to the initial cluster center with the smallest Euclidean distance to obtain an initial cluster; For each data point of the quality file in the data set, the Euclidean distance between it and each initial cluster center is calculated. Euclidean distance is a commonly used distance measurement method that reflects the distance between two data points in space. Then, each data point is assigned to the initial cluster center closest to it, thus forming K initial cluster clusters. In this way, the data points are preliminarily divided into different clusters.

[0058] Step S44: respectively calculating the mean of all data points in the initial clusters, and taking the mean as the new cluster center; For each initial cluster, calculate the mean of all data points in the cluster in each dimension. This mean point will be used as the new cluster center. By calculating the mean, the cluster center can more accurately reflect the characteristics of the data points in the cluster. The new cluster center will be used in the next round of iteration.

[0059] Step S45: Repeat the process of allocating data points and updating cluster centers until the stop condition is met and the final cluster is determined; Repeat steps S43 and S44, continuously assign data points to the nearest cluster center, and update the location of the cluster center. This process will continue until the stop condition is met. The stop condition can be that the cluster center no longer changes, reaches the maximum number of iterations, or reaches a certain error threshold. When the stop condition is met, the algorithm stops iterating and obtains the final cluster.

[0060] Step S46: Determine the importance level corresponding to each final cluster according to the numerical range of the final cluster.

[0061] Finally, according to the numerical range of the clusters finally obtained, the importance level corresponding to each cluster is determined. Different importance levels can be assigned to different numerical ranges according to actual conditions. For example, clusters with larger values ​​can be considered to have high importance levels, and clusters with smaller values ​​can be considered to have low importance levels. In this way, the importance level of quality files is divided.

[0062] Furthermore, in step 41, the normalization process is calculated using the following formula: , In the formula, Y i After normalization, i A sample of usage frequency data; K i For the i A sample of usage frequency data; K max Indicates the maximum value of all usage frequency data samples; K min Indicates the minimum value of all usage frequency data samples.

[0063] Furthermore, in step S43, the Euclidean distance is calculated by the following formula: , In the formula, d ( x, C i ) represents the Euclidean distance function; x is the normalized data object; C i Indicates i Cluster centers; n is the number of samples in the data set; x j Indicates the data set j A data object with a variable; C ij Indicates j The variable i Cluster centers.

[0064] Furthermore, in step S45, the stop condition is specifically that the cluster center no longer changes or reaches a preset update number.

[0065] S5. Classify and store the quality documents according to the importance level, and generate a summary of key quality document information as a label for each quality document; S6. Model application: Obtain the quality files to be executed, extract key quality file information and input it into the random forest model, and output the predicted quality control results; S7. Adjust the quality documents to be executed based on the predicted quality control results.

[0066] Furthermore, in step S5, the quality files are classified and stored according to the importance level, including the following steps: Step S51: Establishing a storage structure: establishing different folder or directory storage structures according to importance levels; The purpose of this step is to better organize and manage quality files. By creating different folders or directories according to the importance level, you can intuitively classify and store quality files. For example, you can create folders such as "high importance level", "medium importance level", and "low importance level". In this way, when searching and accessing quality files, you can quickly locate the corresponding storage location according to its importance level, improving management efficiency.

[0067] Step S52: Automatic classification: Automatically identify the usage frequency of quality documents, determine the importance level according to the usage frequency and store it in the corresponding storage structure; In this step, the system automatically identifies the frequency of use of quality files. The frequency of use can be measured by the indicators mentioned above, such as the number of visitors and the number of modifications. Based on the frequency of use, combined with the importance level classification determined by the K-means clustering method, the quality files are automatically stored in the corresponding importance level folder. For example, if a quality file is used frequently and is judged to be of high importance, it will be automatically stored in the "High Importance Level" folder. This automatic classification method can reduce manual intervention and improve the accuracy and efficiency of storage.

[0068] Step S53: Dynamic monitoring: Set a specific time step and dynamically monitor the importance level through the usage frequency of the stored quality files. When the importance level changes, migrate the quality files to the storage structure corresponding to the importance level and keep the migration log.

[0069] In order to ensure that the storage of quality files is always consistent with their actual importance level, dynamic monitoring is required. A specific time step can be set, such as daily, weekly or monthly. During this time interval, the system will re-evaluate the frequency of use of the stored quality files and determine whether the importance level needs to be adjusted based on the new frequency of use. If the importance level changes, the system will automatically migrate the quality files to the storage structure of the corresponding importance level. At the same time, in order to facilitate traceability and management, the system will keep a migration log of quality files, recording the time of each migration, the source storage location, the target storage location, and the reason for the migration. This can ensure that the storage management of quality files is more flexible and efficient, and adapt to changing business needs.

[0070] Furthermore, in step S5, the generation of a summary of key quality file information as a label for each quality file includes the following steps: 1. Assign numerical values ​​to categorical variables in key quality document information; There may be some categorical variables in the key quality document information, such as the type of quality document, the project stage, etc. In order to include these categorical variables in the summary, a numerical value needs to be assigned to each categorical variable. This numerical value can be determined according to certain rules or coding systems. For example, the numbers 1, 2, and 3 can be used to represent different quality document types, or 0 and 1 can be used to represent different stages of the project. By assigning numerical values, categorical variables can be converted into a computable and sortable form.

[0071] 2. Arrange the values ​​according to the order of importance of the quality file information variables to obtain a numerical sequence summary; After assigning values ​​to the categorical variables, these values ​​need to be arranged in the order of importance of the quality document information variables. This order of importance can be determined by the previous random forest model, with the values ​​corresponding to the variables with higher importance being ranked first. In this way, a series of numerical sequence summaries can be obtained, which can concisely reflect the key features of the quality document. For example, if the key information of a quality document includes the type "product quality standard" (assigned with a value of 1), the importance ranking is high (assuming that high importance corresponds to a value of 2), etc., then the numerical sequence summary after being arranged in order of importance may be "21".

[0072] 3. using the numerical sequence summary as a label of the quality file, and generating annotations for each numerical value of the label; Finally, the obtained numerical sequence summary is used as the label of the quality file. This label can be added to the file name, attribute information or specific field in the storage system of the quality file to facilitate quick identification and management of the quality file. At the same time, in order to better understand the meaning of the label, it is necessary to generate annotations for each numerical value of the label. The annotation is a specific explanation of the categorical variable represented by the numerical value. In this way, when viewing the label of the quality file, the key information of the quality file can be quickly understood through the annotation.

[0073] Furthermore, in step S6, the quality file to be executed is also compared with the key quality file information similarity of the classified and stored quality files to determine the importance level of the quality file to be executed and store it.

[0074] In this embodiment, when there is a quality file to be improved (i.e., a quality file to be executed), in addition to inputting its key quality file information into the random forest model to predict the quality control results according to the previous steps, the key quality file information of the quality file to be executed is also compared for similarity with the key quality file information of the classified and stored quality files.

[0075] First, determine the calculation method for similarity. Common text similarity calculation methods can be used, such as cosine similarity, Jaccard similarity, etc. If the key quality document information is numerical data, methods such as Euclidean distance can also be used.

[0076] Then, the similarity between the quality file to be executed and each quality file that has been classified and stored is calculated one by one. By comparing the similarity values, the stored quality file that is most similar to the quality file to be executed can be found.

[0077] According to the importance level of the most similar stored quality file found, the importance level of the quality file to be executed can be preliminarily determined. If the most similar stored quality file is of high importance level, then the quality file to be executed is likely to have high importance as well.

[0078] Finally, the quality files to be executed are stored according to the determined importance level. The storage method is consistent with the quality files that have been classified and stored, that is, they are placed in folders or storage structures corresponding to the importance level. The advantage of this is that it can make full use of the existing classified storage information, more accurately determine the importance level of the quality files to be executed, and improve the efficiency and accuracy of quality file management.

[0079] Furthermore, in step S7, adjusting the quality file to be executed according to the predicted quality control result includes the following steps: Step S71: When the predicted quality control result is lower than the preset value, check the key quality file information; In this step, if the quality control result of the quality document to be executed is predicted by the random forest model to be lower than the preset value, it means that there may be problems with the quality document and it needs to be adjusted. First, check the key quality document information, which is the information that has a significant impact on the quality control results determined by the previous steps. The purpose of the check is to determine which key information may have led to unsatisfactory quality control results.

[0080] Step S72: adjusting abnormal key quality file information according to quality control requirements; After checking the critical quality document information that may have problems, adjust these abnormal critical information according to the specific needs of quality control. The adjustment methods can include modifying parameter values, updating standards, adjusting processes, etc. For example, if it is found that the product quality standard in a critical quality document is too high and difficult to achieve, it may be necessary to appropriately lower the standard according to the actual situation.

[0081] Step S73: input the adjusted key quality document information into the random forest model to re-predict the quality control results; The adjusted key quality document information is input into the random forest model again to re-predict the quality control results. Through re-prediction, it can be evaluated whether the adjusted quality document can meet the quality control requirements. If the predicted result is still lower than the preset value, it is necessary to continue to adjust.

[0082] Step S74: repeating steps S71 - S73 until the predicted quality control result is not lower than the preset value and the quality file to be executed determines that it meets the execution conditions; Repeat steps S71 to S73 until the predicted quality control result is not less than the preset value. This process may require multiple iterations, and the key quality file information is adjusted and re-predicted in each iteration until the quality control result meets the requirements. When the predicted quality control result meets the preset value, it means that the quality file to be executed meets the execution conditions after adjustment, and the next step can be performed.

[0083] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A multi-level quality file management method based on big data, characterized by: The following steps are involved: S1. Collecting a large-scale sample of quality file data, wherein the quality file data includes quality file information and its corresponding quality control result data; S2. Establish a random forest model between quality document information and quality control results to identify key quality document information that significantly affects quality control results; S3. Real-time monitoring and acquisition of the usage frequency of quality documents; the usage frequency includes the number of people who accessed the quality documents and the number of times they were modified; S4. Clustering the usage frequencies using a cluster analysis method, and dividing the importance levels of the quality files according to the clustering results; S5. Classify and store the quality documents according to the importance level, and generate a summary of key quality document information as a label for each quality document; S6. Model application: Obtain the quality files to be executed, extract key quality file information and input it into the random forest model, and output the predicted quality control results; S7. Adjust the quality documents to be executed based on the predicted quality control results.

2. According to the multi-level quality file management method based on big data according to claim 1, it is characterized by: The quality document information includes quality manual, standard operating procedures, checklists, quality plans, quality records, non-conforming product handling procedures, change control records, risk management documents, training records, and customer feedback and complaint records.

3. The multi-level quality file management method based on big data according to claim 1 is characterized by: The quality control results include non-conforming rate, defect density, inspection pass rate, customer complaint rate, return rate, on-time delivery rate, product life, compliance rate, customer satisfaction and statistical process control data.

4. The multi-level quality file management method based on big data according to claim 1 is characterized by: In step S2, the random forest model includes the following construction steps: S21, divide the data set into a training set and a test set; S22, randomly extracting several quality file information variables from the training set as explanatory variables and quality control results as response variables to generate a decision tree; S23, repeating the operations of steps S21-S22 until all quality file information variables are traversed to obtain a random forest containing large-scale decision trees; S24, calculate the prediction error of each decision tree in the random forest for the test set; S25. Sort the importance of quality document information variables according to the size of the prediction error; S26. Select the top 40% of quality document information variables in importance as key quality document information.

5. The multi-level quality file management method based on big data according to claim 1 is characterized by: In step S4, the cluster analysis method is configured as a K-means clustering method, comprising the following steps: S41, using normalization method to process the frequency data variable so that the data falls within the range of 0 to 1; S42, randomly select K data points from the data set as initial cluster centers; S43, for each data point of the quality file, calculating the Euclidean distance between it and each initial cluster center, and assigning each data point to the initial cluster center with the smallest Euclidean distance, to obtain an initial cluster; S44, respectively calculating the mean of all data points in the initial clusters, and using the mean as a new cluster center; S45, continuously repeating the process of allocating data points and updating cluster centers until a stop condition is met, and determining a final cluster; S46: Determine the importance level corresponding to each final cluster according to the numerical range of the final cluster.

6. The multi-level quality file management method based on big data according to claim 5 is characterized by: In step S45, the stop condition is specifically that the cluster center no longer changes or reaches a preset update number.

7. The multi-level quality file management method based on big data according to claim 1 is characterized by: In step S5, the quality files are classified and stored according to the importance level, including the following steps: S51, establishing storage structure: establishing different folder or directory storage structures according to importance levels; S52, automatic classification: automatically identify the frequency of use of quality documents, determine the importance level according to the frequency of use and store it in the corresponding storage structure; S53, Dynamic Monitoring: Set a specific time step to dynamically monitor the importance level through the usage frequency of the stored quality files. When the importance level changes, migrate the quality files to the storage structure corresponding to the importance level and keep the migration log.

8. The multi-level quality file management method based on big data according to claim 1 is characterized by: In step S5, the generation of a summary of key quality file information as a label for each quality file includes the following steps: Assign numerical values ​​to categorical variables in key quality document information; Arrange the values ​​in the order of importance of the quality file information variables to obtain a numerical sequence summary; The numerical sequence summary is used as a label of the quality file, and annotations are generated for each numerical value of the label.

9. The multi-level quality file management method based on big data according to claim 1 is characterized by: In step S6, the quality file to be executed is also compared with the key quality file information similarity of the classified and stored quality files to determine the importance level of the quality file to be executed and store it.

10. The multi-level quality file management method based on big data according to claim 1, characterized in that: In step S7, adjusting the quality file to be executed according to the predicted quality control result includes the following steps: S71. When the predicted quality control result is lower than the preset value, check the key quality document information; S72. Adjust abnormal key quality document information according to quality control requirements; S73. Input the adjusted key quality document information into the random forest model to re-predict the quality control results; S74. Repeat steps S71-S73 until the predicted quality control result is not lower than the preset value and the quality file to be executed determines that it meets the execution conditions.

Images