现代纺织技术2025，Vol.33Issue(6)：82-89,8.DOI:10.19398/j.att.202410026
基于T-Apriori算法的纺织品质检数据分析
Analysis of textile quality inspection data based on T-Apriori algorithm
摘要
Abstract
Currently,research on association rules for textile quality inspection data is still relatively limited and mostly in the preliminary exploration stage.There are few related studies on improvement schemes for the algorithms adopted in this field,and systematic optimization strategies have not yet been formed.Furthermore,using the traditional Apriori algorithm to mine association rules can be time-consuming in dealing with large datasets,and the determination of rule metric thresholds lacks transparency.Therefore,the purpose of this paper is to address the issue of long computation times when mining association rules from unqualified textile quality inspection data,as well as to solve the problem of the subjective and non-transparent determination of the support threshold. The T-Apriori algorithm optimized based on the traditional Apriori algorithm was adopted.The core idea of this algorithm lies in compressing and storing the Boolean matrix of data in a triplet form.Specifically,each transaction in the Boolean matrix is converted into a set of triplets,where each non-zero element is represented as a triplet(i,j and v),with i and j being the row and column indices,and v being the value of the element.The dataset scanned by the algorithm is the converted triplet set,and the calculations of support,confidence,and lift are also performed using data retrieved from these triplet sets.Unqualified textile quality inspection data are very sparse,with most elements being zero,and triplets only store non-zero elements,thereby effectively reducing storage space and enhancing computational efficiency.For determining the support threshold,the trend in the number of itemsets corresponding to the frequency of candidate 1-itemsets is analyzed to adjust the support threshold,allowing it to better adapt to the characteristics of the data and identify relatively high-frequency itemsets. The experimental results show that the trend in the number of candidate 1-itemsets can be used to identify relatively high-frequent itemsets,and the support threshold for the itemset {brand and total unqualified inspection items} is set to 0.002.The T-Apriori algorithm demonstrates significant performance improvements compared to the traditional Apriori algorithm and its optimized version,C-Apriori.Its runtime is only 40％of that of the traditional Apriori algorithm.As the volume of data increases,the reduction in runtime for the T-Apriori algorithm is even more pronounced,as shown in Fig.5.The lower the support threshold,the larger the difference in runtime between the T-Apriori algorithm and the traditional Apriori algorithm becomes,indicating a more significant reduction in runtime for the T-Apriori algorithm,as illustrated in Fig.6.In summary,the T-Apriori algorithm exhibits superior processing performance in environments with large data volumes and low support thresholds.By mining textile quality inspection data from 2018 to 2023,72 strong association rules are obtained,and based on these rules,two regulatory recommendations are proposed to the supervision department.The adoption of the T-Apriori algorithm greatly improves the analysis efficiency of textile quality inspection data,providing a more efficient data analysis tool for quality supervision and decision support.This has important practical application value.关键词
关联规则/Apriori算法/三元组/质量检测/纺织品Key words
association rules/Apriori algorithm/triplets/quality inspection/textiles分类
轻工纺织引用本文复制引用
吕沿沿,薛文良,魏孟媛,马颜雪..基于T-Apriori算法的纺织品质检数据分析[J].现代纺织技术,2025,33(6):82-89,8.基金项目
国家重点研发计划项目(2022YFF0607203) （2022YFF0607203）
