A method for mining data fluctuation relationship of a silk making workshop based on association rules

Abstract

The application provides a kind of mining method of data fluctuation relationship of silk making workshop based on association rule, and belongs to the field of data mining.For the problem that fluctuation relationship between silk making process parameters is difficult to describe, the application fully mines fluctuation relationship between different variables with the help of association rule.Relying on data discretization, a data preprocessing method for association rule mining of silk making data is designed, and silk making data is converted into fluctuation data containing original data information;Further, the fluctuation rule formula of the data to be mined is designed by using the principle of association rule.The application fully mines the fluctuation rule between different process parameters, improves the interpretability of the rule and the accuracy of the association result, can accurately judge whether there is a fluctuation relationship between each process parameter of the silk making workshop, is conducive to the statistics and management of the process parameters of the silk making workshop, and is convenient for adjusting process parameters to optimize the process when the quality of silk making products is problematic.

Description

Technical Field

[0001] This invention relates to a method for mining data fluctuation relationships in a silk-making workshop based on association rules, belonging to the field of data mining. Background Technology

[0002] With the rapid development of information technologies such as cloud computing, big data, the Internet of Things, and artificial intelligence, and the digital transformation of traditional industries, we have entered the era of explosive data growth—the era of big data. Time-series data mining, similar to that used in silk refining, is a current research hotspot. However, traditional methods are somewhat insufficient in mining the fluctuation relationships between data elements in time-series data, and cannot accurately determine whether there are fluctuation relationships between various process parameters in the silk refining workshop. In industrial big data, the changing relationships between elements are a crucial aspect of data mining. Summary of the Invention

[0003] The purpose of this application is to provide a method for mining data fluctuation relationships in a filament processing workshop based on association rules. Addressing the problem that existing methods cannot accurately and intuitively describe the fluctuation relationships between various process parameters in a filament processing workshop, this method utilizes association rules to fully mine the fluctuation rules of process parameters in the filament processing workshop. Furthermore, the mined rules are highly interpretable, effectively solving the problem that existing methods cannot meet the needs of time-series data association mining and analysis in filament processing workshops.

[0004] The technical solution of this application to achieve the above objectives mainly provides a method for mining the data fluctuation relationship in a silk-making workshop based on association rules, including the following steps:

[0005] Step (1): Determine the process parameters with fluctuation relationship as needed, obtain the corresponding yarn making data, and discretize the data to be mined based on the maximum fluctuation of each process parameter data;

[0006] Step (2): Based on the fluctuations between the data to be mined, the fluctuations between the process parameters are subdivided into four types: positive low fluctuation amplitude - positive and negative high fluctuation amplitude; positive high fluctuation amplitude - positive and negative strong fluctuation amplitude; negative low fluctuation amplitude - positive and negative high fluctuation amplitude; negative high fluctuation amplitude - positive and negative strong fluctuation amplitude.

[0007] Step (3): Design a support calculation formula based on filament production data and data fluctuation type, and determine the frequent itemsets in the process parameters;

[0008] Step (4): Design a confidence calculation formula based on silk production data and data fluctuation type;

[0009] Step (5): Based on the results calculated using the formulas for support and confidence, and the set minimum support threshold and minimum confidence threshold, determine whether there is a fluctuation relationship between the process parameters.

[0010] Specifically, the discretization process in step (1) includes the following steps:

[0011] Step (1.1): Iterate through the differences between the values ​​of each obtained yarn production data at all times and the values ​​at the previous time, and divide the data into several equal parts based on the maximum absolute value of the differences.

[0012] Step (1.2): Reassign values ​​according to the absolute value of each difference in each equal interval. Data values ​​in the same interval are the same. Specifically, assign values ​​to each equal interval in sequence. All values ​​in the first interval are 1, all values ​​in the second interval are 2, and so on.

[0013] Step (1.3): If the value of the yarn production data at a certain moment increases compared to the previous moment, add "+" before the reassigned data; if it decreases compared to the previous moment, add "-" before the reassigned data; if the value of the yarn production data at a certain moment is the same as the previous moment, reassign all values ​​to "0".

[0014] Specifically, the steps of step (2) are as follows:

[0015] Analyzing the fluctuation relationship between two parameters, only fluctuation relationships where a change in one parameter at a certain moment simultaneously causes a valid change in the other parameter can be used as fluctuation relationship rules in association rules;

[0016] When the intervals for both parameters A and B are α, the value of parameter A at time β is X, and the value of parameter B is Y.

[0017] ①When hour, That is, when the positive fluctuation amplitude of parameter A at time β is small, it causes the positive and negative fluctuation amplitude of parameter B to be large;

[0018] ②When When (X≤Y≤α)∩(-α≤Y≤-X), that is, when the positive fluctuation amplitude of parameter A is large at time β, it causes stronger positive and negative fluctuations of parameter B.

[0019] ③When hour, That is, when the negative fluctuation amplitude of parameter A at time β is small, it causes the positive and negative fluctuation amplitudes of parameter B to be large.

[0020] ④ When hour, That is, when the negative fluctuation amplitude of parameter A is large at time β, it causes stronger positive and negative fluctuations in parameter B.

[0021] Specifically, the formula for calculating the support between the process parameters designed in step (3) is as follows:

[0022] S(A,B)=s(AB)+s(BA)

[0023] in:

[0024]

[0025] S(A,B) represents the total support of parameters A and B in the yarn-making workshop. s(AB) represents the support of parameters A and B when using parameters A as the fluctuation benchmark by mining second-order frequent itemsets and performing frequent item mining on parameters A and the corresponding data at a certain time. Similarly, s(BA) represents the support of parameters A and B when using parameters B as the fluctuation benchmark by mining second-order frequent itemsets and performing frequent item mining on parameters B and the corresponding data at a certain time. This indicates the support level when a small positive fluctuation in parameter A leads to a large positive or negative fluctuation in parameter B, and so on.

[0026] Specifically, the formula for calculating the confidence level between the process parameters designed in step (4) is as follows:

[0027]

[0028] Where C(A->B) represents the total confidence level of parameter A pointing to parameter B. This represents the confidence level when a small positive fluctuation in parameter A leads to a large positive or negative fluctuation in parameter B, and so on; N represents all categories of values ​​for parameter A that satisfy the four fluctuation conditions.

[0029] Specifically, the parameter fluctuation results obtained in step (5) are as follows:

[0030] Step (5.1) sets a minimum support threshold and a minimum confidence threshold for each process parameter whose fluctuation relationship needs to be determined in step (1). For each process parameter, if the value calculated by the support formula is higher than the set minimum support threshold, then this process parameter is a frequent term.

[0031] Step (5.2) involves obtaining all the frequent itemsets, which together form a frequent itemset.

[0032] Step (5.3) Extract pairs of frequent items from the frequent item set and calculate the confidence level using the confidence level formula. When the calculated result is higher than the set minimum confidence level threshold, the pair of process parameters have a high confidence level fluctuation rule.

[0033] The beneficial effects of this invention are:

[0034] 1. After data discretization, correlation rules and different fluctuation types are used to mine the fluctuation correlation rules between silk production data.

[0035] 2. Based on the data from the silk-making workshop, we designed support and confidence formulas, which eliminated rules with no correlation between elements and rules with low fluctuation intensity, while improving the interpretability of the rules and the accuracy of the correlation results.

[0036] 3. When quality problems occur in the silk-making products, the process can be optimized by adjusting the process parameters. Attached Figure Description

[0037] Figure 1 This is a flowchart of a method for mining data fluctuation relationships in a silk-making workshop based on association rules, as described in this invention.

[0038] Figure 2 This is a data display of the silk-making part described in this invention;

[0039] Figure 3 This is a display of the discretized data of the filament-making process as described in this invention;

[0040] Figure 4 This is a comparison test result diagram of the data fluctuation relationship mining method of the silk-making workshop based on association rules described in this invention and other methods. Detailed Implementation

[0041] The present invention will be further described below with reference to the accompanying drawings and embodiments, but the scope of the present invention is not limited thereto.

[0042] Example 1: As Figure 1-4 As shown, a method for mining data fluctuation relationships in a silk refining workshop based on association rules mainly includes the following steps:

[0043] Step (1): Determine the process parameters with fluctuation relationship as needed, obtain the corresponding yarn making data, and discretize the data to be mined based on the maximum fluctuation of each process parameter data;

[0044] Step (2): Based on the fluctuations between the data to be mined, the fluctuations between the process parameters are subdivided into four types: positive low fluctuation amplitude - positive and negative high fluctuation amplitude; positive high fluctuation amplitude - positive and negative strong fluctuation amplitude; negative low fluctuation amplitude - positive and negative high fluctuation amplitude; negative high fluctuation amplitude - positive and negative strong fluctuation amplitude.

[0045] Step (3): Design a support calculation formula based on filament data and data fluctuation type. Since the fluctuation between process parameters is relative, that is, the fluctuation of parameter A to parameter B and parameter B to parameter A are different, design a support formula to determine the frequent itemsets in the process parameters.

[0046] Step (4): Design a confidence calculation formula based on silk production data and data fluctuation type;

[0047] Step (5): Based on the results calculated using the formulas for support and confidence, and the set minimum support threshold and minimum confidence threshold, determine whether there is a fluctuation relationship between the process parameters.

[0048] Furthermore, the specific steps for discretization in step (1) include:

[0049] Step (1.1): The silk-making data is time-series data, and the manufacturing process involves many processes. In this application, the proportioning and flavoring processes are used as examples. Figure 2 As shown, this involves 11 processing parameters and 1 quality indicator. The processing parameters include the instantaneous blending ratio of stems and filaments, the instantaneous blending accuracy of stems and filaments, the instantaneous blending ratio of airflow filaments, the instantaneous blending accuracy of airflow filaments, the cumulative amount of thin-plate filaments, the cumulative amount of stems and filaments, and the cumulative amount of fragrance, the cumulative amount of material, the instantaneous fragrance ratio, and the instantaneous fragrance accuracy in the fragrance process. The quality indicator is the output moisture content in the fragrance process. Taking the instantaneous blending ratio of stems and filaments as an example, the difference between the value of each filament processing data at all times and the value at the previous time is iterated. The difference is divided into several equal parts based on the maximum absolute value of the difference. That is, the equal groups are formed according to the specific problem being analyzed. In the implementation case, it is divided into 10 equal parts based on the maximum absolute value of the difference.

[0050] Step (1.2): Reassign values ​​according to the absolute value of each difference in each equal interval. Data values ​​in the same interval are the same. Specifically, assign values ​​to each equal interval in sequence, that is, all values ​​in the first interval are 1, all values ​​in the second interval are 2, and so on.

[0051] Step (1.3): If the value of the yarn-making data at a certain moment increases compared to the previous moment, add a "+" before the reassigned data; if it decreases compared to the previous moment, add a "-" before the reassigned data; if the value of the yarn-making data at a certain moment is the same as the previous moment, reassign all values ​​to "0". The final discretization result is as follows: Figure 3 As shown.

[0052] Furthermore, the specific steps of step (2) are as follows:

[0053] Analyzing the fluctuation relationship between two parameters, only fluctuation relationships where a change in one parameter at a certain moment simultaneously causes a valid change in the other parameter can be used as fluctuation relationship rules in association rules;

[0054] When the intervals for both parameters A and B are α, the value of parameter A at time β is X, and the value of parameter B is Y.

[0055] ①When hour, That is, when the positive fluctuation amplitude of parameter A at time β is small, it causes the positive and negative fluctuation amplitude of parameter B to be large;

[0056] ②When When (X≤Y≤α)∩(-α≤Y≤-X), that is, when the positive fluctuation amplitude of parameter A is large at time β, it causes stronger positive and negative fluctuations of parameter B.

[0057] ③When hour, That is, when the negative fluctuation amplitude of parameter A at time β is small, it causes the positive and negative fluctuation amplitudes of parameter B to be large.

[0058] ④ When hour, That is, when the negative fluctuation amplitude of parameter A is large at time β, it causes stronger positive and negative fluctuations in parameter B.

[0059] Furthermore, the formula for calculating the support between the process parameters designed in step (3) is as follows:

[0060] S(A,B)=s(AB)+s(BA)

[0061] in:

[0062]

[0063] S(A,B) represents the total support of parameters A and B in the yarn-making workshop. s(AB) represents the support of parameters A and B when using parameters A as the fluctuation benchmark by mining second-order frequent itemsets and performing frequent item mining on parameters A and the corresponding data at a certain time. Similarly, s(BA) represents the support of parameters A and B when using parameters B as the fluctuation benchmark by mining second-order frequent itemsets and performing frequent item mining on parameters B and the corresponding data at a certain time. This represents the support level when a small positive fluctuation in parameter A leads to a large positive or negative fluctuation in parameter B, and so on. Specifically, the mining method limits the order of the maximum frequent item to 2, that is, it limits the order of parameter network mining to 2, which means that the support and confidence levels between every two parameters are calculated.

[0064] Furthermore, the formula for calculating the confidence level between the process parameters designed in step (4) is as follows:

[0065]

[0066] Where C(A->B) represents the total confidence level of parameter A pointing to parameter B. This represents the confidence level when a small positive fluctuation in parameter A leads to a large positive or negative fluctuation in parameter B, and so on; N represents all categories of values ​​for parameter A that satisfy the four fluctuation conditions.

[0067] Furthermore, the parameter fluctuation results obtained in step (5) are as follows:

[0068] Step (5.1) sets a minimum support threshold and a minimum confidence threshold for each process parameter whose fluctuation relationship needs to be determined in step (1). For each process parameter, if the value calculated by the support formula is higher than the set minimum support threshold, then this process parameter is a frequent term.

[0069] Step (5.2) involves obtaining all the frequent itemsets, which together form a frequent itemset.

[0070] Step (5.3) Extract pairs of frequent items from the frequent item set and calculate the confidence level using the confidence level formula. When the calculated result is higher than the set minimum confidence level threshold, the pair of process parameters have a high confidence level fluctuation rule.

[0071] See Figure 4 This image shows the comparative test results of the association rule mining method for silk production data described in this invention. The association rule mining method provided in this application is compared with traditional association rule algorithms FP-growth and Apriori. The mining method in this application is named the UNDULA algorithm, and the parameter settings for each algorithm are as follows:

[0072] (1) UNDULA: S min =5% (where S) (AB)min =2.5%, S (BA)min =2.5%), C min =75%;

[0073] (2) Apriori: S min =5%, C min =75%;

[0074] (3) FP-growth: S min =5%, C min =75%;

[0075] Among them, S min C represents the minimum support threshold. min S represents the minimum confidence threshold. (AB)min S represents the minimum support threshold between parameters A and B, with parameter A as the fluctuation benchmark between data points. (BA)min This represents the minimum support threshold between parameters A and B, with parameter A serving as the fluctuation benchmark between data points.

[0076] Under the same conditions, using a database containing 10,000 data entries, take S... min =5%, C min =75% of the calculated association rules are filtered by Figure 4 It can be seen that, with smaller datasets, the three algorithms mined roughly the same number of rules. However, as the data size increases, the UNDULA algorithm, with the same parameters, can obtain more interpretable association rules compared to the other two algorithms.

[0077] This invention fully explores the fluctuation rules between different process parameters, improves the interpretability of the rules and the accuracy of the correlation results, and can accurately determine whether there are fluctuation relationships between various process parameters in the yarn making workshop. This is beneficial for the statistics and management of process parameters in the yarn making workshop, and makes it easier to optimize the process by adjusting process parameters when problems occur in the quality of yarn products.

[0078] The specific embodiments of the present invention have been described in detail above with reference to the accompanying drawings. The above embodiments are only examples of the present invention and are not intended to limit the implementation and scope of the present invention. All technical solutions that are the same as or equivalent to the contents described in the claims of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for mining data fluctuation relationships in a silk refining workshop based on association rules, characterized in that: Includes the following steps: Step (1): Determine the process parameters with fluctuation relationship as needed, obtain the corresponding yarn making data, and discretize the data to be mined based on the maximum fluctuation of each process parameter data. Step (2): Based on the fluctuations between the data to be mined, the fluctuations between the process parameters are subdivided into four types: positive low fluctuation amplitude - positive and negative high fluctuation amplitude; positive high fluctuation amplitude - positive and negative strong fluctuation amplitude; negative low fluctuation amplitude - positive and negative high fluctuation amplitude; negative high fluctuation amplitude - positive and negative strong fluctuation amplitude. Step (3): Design a support calculation formula based on filament production data and data fluctuation type, and determine the frequent itemsets in the process parameters; Step (4): Design a confidence calculation formula based on silk production data and data fluctuation type; Step (5): Based on the results calculated using the formulas for support and confidence, and the set minimum support threshold and minimum confidence threshold, determine whether there is a fluctuation relationship between the process parameters; The specific steps of step (2) are as follows: Analyzing the fluctuation relationship between two parameters, only fluctuation relationships where a change in one parameter at a certain moment simultaneously causes a valid change in the other parameter can be used as fluctuation relationship rules in association rules; When the intervals for both parameters A and B are α, the value of parameter A at time β is X, and the value of parameter B is Y. ① When hour, That is, when the positive fluctuation amplitude of parameter A at time β is small, it causes the positive and negative fluctuation amplitude of parameter B to be large. When , , i.e. when the positive fluctuation amplitude of parameter A at time point β is larger, it causes stronger positive and negative fluctuations of parameter B. ③ When hour, That is, when the negative fluctuation amplitude of parameter A at time β is small, it causes the positive and negative fluctuation amplitude of parameter B to be large. When , , when the negative fluctuation amplitude of parameter A is larger at time point β, it causes stronger positive and negative fluctuations of parameter B.

2. The mining method of data fluctuation relations of a silk making plant based on association rules according to claim 1, characterized in that: The specific steps for discretization in step (1) include: Step (1.1): Iterate through the differences between the values ​​of each obtained yarn production data at all times and the values ​​at the previous time, and divide the data into several equal parts based on the maximum absolute value of the differences. Step (1.2): Reassign values ​​according to the absolute value of each difference in each equal interval. Data values ​​in the same interval are the same. Specifically, assign values ​​to each equal interval in sequence. All values ​​in the first interval are 1, all values ​​in the second interval are 2, and so on. Step (1.3): If the value of the yarn-making data at a certain moment increases compared to the previous moment, add a "+" sign before the reassigned data; if it decreases compared to the previous moment, add a "+" sign before the reassigned data. If the value of the yarn production data at a certain moment is the same as that at the previous moment, then all values ​​are reassigned to "0".

3. The method of claim 1, wherein the association rule-based mining of data fluctuation relationships in a silk filament production plant is characterized by, The formula for calculating the support between the process parameters designed in step (3) is as follows: ； in: ； S(A, B) represents the total support of parameters A and B in the yarn processing workshop. s(AB) represents the support of parameters A and B when using parameter A as the fluctuation benchmark, obtained by mining second-order frequent itemsets and performing frequent item mining on parameter A and the corresponding data at a specific time. Similarly, s(BA) represents the support of parameters A and B when using parameter B as the fluctuation benchmark, obtained by mining second-order frequent itemsets and performing frequent item mining on parameter B and the corresponding data at a specific time. This indicates the support level when a small positive fluctuation in parameter A leads to a large positive or negative fluctuation in parameter B, and so on.

4. The method of claim 1, wherein the association rule-based mining of data fluctuation relationships in a cut tobacco plant is characterized by, The formula for calculating the confidence level between the process parameters designed in step (4) is as follows: ； ； in This indicates the total confidence level of parameter A pointing to parameter B. This represents the confidence level when a small positive fluctuation in parameter A leads to a large positive or negative fluctuation in parameter B, and so on; N represents all categories of values ​​for parameter A that satisfy the four fluctuation conditions.

5. The method of claim 1, wherein the association rule-based mining of data fluctuation relationships in a silk filament production plant is characterized by, The parameter fluctuation results obtained in step (5) are as follows: Step (5.1) sets a minimum support threshold and a minimum confidence threshold for each process parameter whose fluctuation relationship needs to be determined in step (1). For each process parameter, if the value calculated by the support formula is higher than the set minimum support threshold, then this process parameter is a frequent term. Step (5.2) All the frequent items obtained form a frequent itemset; Step (5.3) Extract pairs of frequent items from the frequent item set and calculate the confidence level using the confidence level formula. When the calculated result is higher than the set minimum confidence level threshold, the pair of process parameters have a high confidence level fluctuation rule.

Images