Fabric anti-static detection data intelligent classification method

Abstract

The invention relates to the technical field of data processing, in particular to a fabric anti-static detection data intelligent classification method, which comprises the following steps: acquiring the divisible degree of each item of data in each subset according to the non-uniformity degree and the frequency of selecting each item of data as a division feature; obtaining the adjusted Gini impurity of each item of data in each subset according to the segmentation degree; and classifying the anti-static data of the fabric based on the adjusted Gini impurity. According to the invention, the classification of the anti-static detection data of the fabric is more accurate.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and more specifically, to an intelligent classification method for fabric anti-static detection data. Background Art

[0002] In the field of modern fabric manufacturing and application, especially in the semiconductor, electronic equipment assembly and medical industries, electrostatic discharge (ESD) may cause great damage, leading to equipment failure and errors in the production process, so the demand for anti-static fabrics is becoming more and more urgent; CART (Classification and Regression Tree) algorithm is widely used in the classification of anti-static test data and has become an important tool for evaluating the anti-static performance of fabrics. By analyzing the anti-static performance data of different fabrics, the CART algorithm can efficiently build a decision tree model to identify and classify the performance differences of various fabrics.

[0003] The patent application document with the publication number CN115982722A discloses a vulnerability classification detection method based on a decision tree. It first uses Python to crawl vulnerability data information in the vulnerability database and integrates it into an original data set; then, it selects data features and divides the data set, optimizes the divided data set, determines the classification order, and sets the GINI coefficient threshold to establish a CART decision tree; then, it conducts a reasonable evaluation of the constructed model and evaluates the performance of the constructed model based on accuracy, error rate and running time.

[0004] In the process of classifying the anti-static data of fabrics through the CART algorithm of Gini impurity, since the value range of each data in the anti-static data of fabrics is different, directly classifying the data according to the CART algorithm of Gini impurity will make it more inclined to select data items containing more data for classification, while ignoring the internal performance of each data, resulting in inaccurate classification. Summary of the invention

[0005] In order to solve the problem that the CART algorithm tends to select features with more values ​​for classification, ignores the performance of each data in each subtree during the classification process, and cannot select the most suitable classification feature for each subtree for classification, so that the classification of data may be biased, the present invention proposes an intelligent classification method for fabric anti-static detection data, which includes the following steps: Obtain anti-static data of several batches of fabrics; divide all items of anti-static data of all batches of fabrics into subsets to obtain several subsets, as well as the Gini impurity of each item of data in each subset; obtain the unevenness of each item of data in each subset: , Indicates the unevenness of the j-th data in the m-th subset; represents the number of all j-th data items in the m-th subset; N represents the mean of the number of all data items in all subsets; Represents the j-th data of the b+1-th batch of fabrics in the j-th data arrangement sequence of the m-th subset; Represents the j-th data of the b-th batch of fabrics in the j-th data arrangement sequence of the m-th subset; Represents the range of the j-th data in all subsets; Represents the range of the jth data item in the mth subset; Get the degree of divisibility of each item in each subset: , Indicates the degree of divisibility of the j-th data item in the m-th subset; Indicates the number of times the j-th data item is selected as the partition feature during the partition process; represents the partitioning parameter; Indicates the unevenness of the division of the j-th data; obtain the adjusted Gini impurity of each data in each subset: , represents the adjusted Gini impurity of the j-th data in the m-th subset; Represents the Gini impurity of the j-th data in the m-th subset; the anti-static data of the fabric is classified based on the adjusted Gini impurity.

[0006] The innovation of the present invention lies in obtaining the unevenness of each data in each subset and the number of times each data is selected as a division feature according to the internal performance differences of each data in the fabric anti-static data, and then obtaining the divisibility of each data in each subset; obtaining the adjusted Gini impurity of each data in each subset according to the divisibility; and classifying the fabric anti-static data based on the adjusted Gini impurity, thereby making the classification of the fabric anti-static detection data more accurate.

[0007] Preferably, the method of performing subset division on all items of data in the antistatic data of all batches of fabrics to obtain several subsets, and the Gini impurity of each item of data in each subset, includes: Preset a partition parameter , record all items of anti-static data of all batches of fabrics as the initial subset, divide the initial subset into two subsets, and obtain the Gini impurity of each item of data in each subset; obtain the first mean of each subset; record the subset with the smallest first mean as the first subset to be divided; divide the first subset to be divided into two subsets; and so on, divide the total times, and obtain several subsets and the Gini impurity of each data in each subset.

[0008] By calculating the Gini impurity of the anti-static data of all batches of fabrics, the partitioning strategy can be dynamically adjusted to ensure that the heterogeneity of the anti-static effects of the divided subsets is maximized.

[0009] Preferably, obtaining the first mean of each subset includes: The mean of the Gini impurity of all items in each subset is recorded as the first mean of each subset.

[0010] Preferably, obtaining the j-th data arrangement sequence of the m-th subset includes: Sort the j-th data of the fabric in the m-th subset from small to large to obtain the arrangement sequence of the j-th data in the m-th subset.

[0011] Preferably, the acquisition of the division unevenness includes: ; In the formula, Represents the mean of the range of the j-th data in all subsets; Represents the cumulative sum of the ranges of the j-th data in all subsets; Indicates the unevenness of the division of the j-th data.

[0012] When the data is divided unevenly, the decision tree is able to make finer divisions on these features, thereby reducing the misclassification rate of the model.

[0013] Preferably, the classifying the anti-static data of the fabric based on the adjusted Gini impurity includes: The minimum value of the adjusted Gini impurity is used as the optimal feature selection in the CART decision tree algorithm for subset division, and the final decision tree is constructed. The final decision tree is used to classify the anti-static data of all batches of fabrics.

[0014] The adjusted Gini impurity makes the classification selection more reasonable by introducing a correction method, reduces the misclassification rate in the model, and thus improves the classification effect of anti-static performance.

[0015] Preferably, the obtaining of anti-static data of several batches of fabrics includes: Sampling is performed on multiple batches of fabrics. For any batch of fabrics, the fiber diameter of the batch of fabrics is measured using an electron microscope, the fabric thickness of the batch of fabrics is measured using a digital thickness gauge, the batch of fabrics is placed on a resistance tester, and the batch of fabrics is measured to obtain the point-to-point resistance of the batch of fabrics; the fiber diameter, fabric thickness and point-to-point resistance of the batch of fabrics are sequentially used as the anti-static data of the batch of fabrics.

[0016] The present invention has the following beneficial effects: since the different value ranges of each data in the anti-static data of fabrics lead to large differences in the internal performance of each data, the present invention obtains the divisibility of each data in each subset according to the degree of unevenness and the number of times each data is selected as a division feature; obtains the adjusted Gini impurity of each data in each subset according to the divisibility; and classifies the anti-static data of fabrics based on the adjusted Gini impurity, thereby making the classification of anti-static detection data of fabrics more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein: Figure 1 The present invention is a flowchart of the steps of an intelligent classification method for fabric anti-static detection data according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0019] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0020] See also Figure 1 , which shows a flowchart of a method for intelligent classification of fabric anti-static detection data provided by an embodiment of the present invention, the method comprising the following steps: S001. Obtain anti-static data for several batches of fabrics.

[0021] It should be noted that the fiber diameter of the fabric directly affects the electrical conductivity and electrostatic properties of the fabric. A smaller fiber diameter usually increases the conductivity of the fabric. The thickness of the fabric affects the electrical conductivity and static electricity accumulation of the fabric. Thinner fabrics may accumulate static electricity more easily. Point-to-point resistance directly reflects the conductivity of the fabric and is an important indicator for evaluating the anti-static performance of the fabric. The resistance of anti-static fabric should be lower than a certain standard. Therefore, when collecting anti-static test data for fabrics, especially when it comes to parameters such as fiber diameter, fabric thickness, and point-to-point resistance, it is usually necessary to use professional instruments and equipment for precise measurement.

[0022] In an embodiment of the present invention, multiple batches of fabrics are sampled, and for any batch of fabrics, the fiber diameter of the batch of fabrics is measured using an electron microscope, the fabric thickness of the batch of fabrics is measured using a digital thickness gauge, the batch of fabrics is placed on a resistance tester, and the batch of fabrics is measured to obtain the point-to-point resistance of the batch of fabrics; the fiber diameter, fabric thickness and point-to-point resistance of the batch of fabrics are sequentially used as the anti-static data of the batch of fabrics.

[0023] S002. Divide all items of data in the anti-static data of all batches of fabrics into subsets to obtain several subsets and the Gini impurity of each item of data in each subset; obtain the unevenness of each item of data in each subset according to the difference in data values ​​of each item of data in each subset; obtain the divisibility of each item of data in each subset according to the unevenness and the number of times each item of data is selected as a dividing feature; obtain the adjusted Gini impurity of each item of data in each subset according to the divisibility.

[0024] It should be noted that in the process of classifying the anti-static data of fabrics through the CART algorithm of Gini impurity, since the value range of each data in the anti-static data of fabrics is different, directly classifying the data according to the CART algorithm of Gini impurity will make it more inclined to select data items containing more data for classification, while ignoring the internal performance of each data. Therefore, the anti-static data of fabrics is divided into subsets, and the internal performance of each data in each subset is analyzed to obtain the divisibility of each data in each subset, and the segmentation factor is obtained according to the divisibility of each data in the subset, and the Gini impurity is adjusted according to the segmentation factor, and the decision tree is divided according to the adjusted Gini impurity.

[0025] In the embodiment of the present invention, a partition parameter is preset. , wherein this embodiment is based on This example is described as an example, and this embodiment is not specifically limited. Depends on the specific implementation situation; All items of data in the anti-static data of all batches of fabrics are recorded as the initial subset, and the initial subset is divided into two subsets using the CART decision tree algorithm, and the Gini impurity of each item of data in each subset is obtained; the mean of the Gini impurity of all items in the subset is recorded as the first mean of the subset; in the two subsets, the subset with the smallest first mean is recorded as the first subset to be divided; the first subset to be divided is divided into two subsets; in the three subsets, the subset with the smallest first mean is recorded as the second subset to be divided, and the second subset to be divided is divided into two subsets; in the four subsets, the subset with the smallest first mean is recorded as the third subset to be divided, and the third subset to be divided is divided into two subsets, and so on, a total of times, and obtain several subsets and the Gini impurity of each data in each subset.

[0026] Each subset includes different items of data, and the CART algorithm of Gini impurity is a prior art, which will not be described in detail in this embodiment.

[0027] 1. Obtain the degree of unevenness of each data item in each subset.

[0028] It should be noted that since the value range and data accuracy of each data are different, there are naturally certain priority differences in the selection of each data for division. Such differences will cause certain interference to the selection of division features, resulting in inaccurate data division; among them, for any data in any subset, the more uneven the data distribution of the data is, the more the data should be divided into multiple categories, so the data segmentation can be judged according to the degree of unevenness of each data in each subset.

[0029] In the embodiment of the present invention, the j-th data of the fabric in the m-th subset is sorted from small to large to obtain the j-th data arrangement sequence of the m-th subset; then the calculation formula for the unevenness of the j-th data in the m-th subset is: ; In the formula, Indicates the unevenness of the j-th data in the m-th subset; represents the number of all j-th data items in the m-th subset; N represents the mean of the number of all data items in all subsets; Represents the j-th data of the b+1-th batch of fabrics in the j-th data arrangement sequence of the m-th subset; Represents the j-th data of the b-th batch of fabrics in the j-th data arrangement sequence of the m-th subset; Represents the range of the j-th data in all subsets; Represents the range of the j-th data item in the m-th subset.

[0030] It should be noted that The larger the value is, the smaller the amount of data in the mth subset is. This type of subset should be considered relatively uniform even if the data is not uniform enough, so that the subset containing less data will not be over-divided. It represents the difference between the difference between two adjacent j-th data items in the j-th data arrangement sequence of the m-th subset and the average difference. The larger the value is, the more uniform the arrangement sequence of the j-th data item in the m-th subset is, and the greater the unevenness of the j-th data item in the m-th subset is.

[0031] At this point, the degree of unevenness of each item of data in each subset is obtained.

[0032] 2. Obtain the degree of divisibility of each data item in each subset.

[0033] It should be noted that in addition to the degree of unevenness, other factors of each data item should also be considered, that is, the more data the data item contains, the more it should be segmented, and the more times the data item is selected in the segmentation process, the lower its priority in subsequent selection should be.

[0034] In the embodiment of the present invention, the calculation formula for obtaining the divisibility of the j-th item of data in the m-th subset is: ; In the formula, Indicates the degree of divisibility of the j-th data item in the m-th subset; Indicates the unevenness of the j-th data in the m-th subset; Indicates the number of times the j-th data item is selected as the partition feature during the partition process; represents the partitioning parameter; Represents the mean of the range of the j-th data in all subsets; Represents the cumulative sum of the ranges of the j-th data in all subsets.

[0035] It should be noted that It indicates the unevenness of the division of the j-th data. For the j-th data, if it cannot be divided well in the previous division processes, the dividing line of the j-th data in the two subsets will not be obvious enough. Therefore, the unevenness of the division of the j-th data is obtained by summing the range of the j-th data in all subsets and the range of the j-th data.

[0036] At this point, the divisibility of each item of data in each subset is obtained.

[0037] 3. Get the adjusted Gini impurity of each data in each subset.

[0038] It should be noted that since the degree of divisibility takes into account the unevenness of each data item in each subset, the preference degree of each data item is optimized from the data itself, which eliminates the influence of the value range of each data item on data selection to a certain extent, making the selection of partitioning features more accurate.

[0039] In the embodiment of the present invention, the calculation formula for obtaining the adjusted Gini impurity of the j-th item of data in the m-th subset is: ; In the formula, represents the adjusted Gini impurity of the j-th data in the m-th subset; Represents the Gini impurity of the j-th data in the m-th subset; Indicates the degree of divisibility of the j-th data item in the m-th subset.

[0040] At this point, the adjusted Gini impurity of each data in each subset is obtained.

[0041] S003. Classify the anti-static data of fabrics based on the adjusted Gini impurity.

[0042] In an embodiment of the present invention, the minimum value of the adjusted Gini impurity is used as the optimal feature selection in the CART decision tree algorithm for subset division, a final decision tree is constructed, and the anti-static data of all batches of fabrics are classified using the final decision tree.

[0043] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for intelligent classification of fabric anti-static detection data, characterized in that: include: Obtain anti-static data for several batches of fabrics; Divide all items of antistatic data of all batches of fabrics into subsets to obtain several subsets and the Gini impurity of each item of data in each subset; obtain the unevenness of each item of data in each subset: , Indicates the unevenness of the j-th data in the m-th subset; represents the number of all j-th data items in the m-th subset; N represents the mean of the number of all data items in all subsets; Represents the j-th data of the b+1-th batch of fabrics in the j-th data arrangement sequence of the m-th subset; Represents the j-th data of the b-th batch of fabrics in the j-th data arrangement sequence of the m-th subset; Represents the range of the j-th data in all subsets; Represents the range of the jth data item in the mth subset; Get the degree of divisibility of each item in each subset: , Indicates the degree of divisibility of the j-th data item in the m-th subset; Indicates the number of times the j-th data item is selected as the partition feature during the partition process; represents the partitioning parameter; Indicates the unevenness of the division of the j-th data; obtain the adjusted Gini impurity of each data in each subset: , represents the adjusted Gini impurity of the j-th data in the m-th subset; Represents the Gini impurity of the j-th data in the m-th subset; the anti-static data of the fabric is classified based on the adjusted Gini impurity.

2. The intelligent classification method for fabric anti-static detection data according to claim 1 is characterized in that: The method of performing subset division on all items of antistatic data of all batches of fabrics to obtain several subsets and the Gini impurity of each item of data in each subset includes: Preset a partition parameter , record all items of anti-static data of all batches of fabrics as the initial subset, divide the initial subset into two subsets, and obtain the Gini impurity of each item of data in each subset; obtain the first mean of each subset; record the subset with the smallest first mean as the first subset to be divided; divide the first subset to be divided into two subsets; and so on, divide the total times, and obtain several subsets and the Gini impurity of each data in each subset.

3. The intelligent classification method for fabric anti-static detection data according to claim 2 is characterized in that: The obtaining of the first mean of each subset includes: The mean of the Gini impurity of all items in each subset is recorded as the first mean of each subset.

4. The intelligent classification method for fabric anti-static detection data according to claim 1 is characterized in that: The acquisition of the j-th data arrangement sequence of the m-th subset includes: Sort the j-th data of the fabric in the m-th subset from small to large to obtain the arrangement sequence of the j-th data in the m-th subset.

5. The intelligent classification method for fabric anti-static detection data according to claim 1 is characterized in that: The acquisition of the division unevenness includes: ； In the formula, Represents the mean of the range of the j-th data in all subsets; Represents the cumulative sum of the ranges of the j-th data in all subsets; Indicates the unevenness of the division of the j-th data.

6. The intelligent classification method for fabric anti-static detection data according to claim 1 is characterized in that: The anti-static data of fabrics are classified based on the adjusted Gini impurity, including: The minimum value of the adjusted Gini impurity is used as the optimal feature selection in the CART decision tree algorithm for subset division, and the final decision tree is constructed. The final decision tree is used to classify the anti-static data of all batches of fabrics.

7. The intelligent classification method for fabric anti-static detection data according to claim 1 is characterized in that: The method of obtaining anti-static data of several batches of fabrics includes: Sampling is performed on multiple batches of fabrics. For any batch of fabrics, the fiber diameter of the batch of fabrics is measured using an electron microscope, the fabric thickness of the batch of fabrics is measured using a digital thickness gauge, the batch of fabrics is placed on a resistance tester, and the batch of fabrics is measured to obtain the point-to-point resistance of the batch of fabrics; the fiber diameter, fabric thickness and point-to-point resistance of the batch of fabrics are sequentially used as the anti-static data of the batch of fabrics.

Images