信息工程与人工智能学院知识库

库存物资的合理分类对企业库存管理活动至关重要，对于每类物资根据其性质有针对性地制定相应的库存控制策略，可以降低库存成本，优化库存管理活动，其中ABC库存分类法是最基础的方法。现有ABC分类研究中大多数仅考虑独立需求物资，没有考虑物资之间内在的关联性。频繁项集挖掘算法产生的频繁模式包含着物资间的内在联系，可以用来优化ABC库存分类，但是挖掘出的频繁模式数量巨大，会导致某些项重复出现，在库存分类调整时可能会出现分类矛盾的问题，而利用闭频繁项集压缩无损的特点就能很好地解决这一问题。闭频繁项集是相关频繁项集的精简表示方式，闭频繁项集挖掘极大减少了挖掘结果中的频繁项集数量，成为近年来数据挖掘领域的一个重要研究课题。当前多个闭频繁项集挖掘算法已经提出，均可应用于ABC库存分类优化问题。DCI_Closed算法是一个经典的闭频繁项集挖掘算法，经过分析发现，其剪枝策略等方面仍有改进空间，算法效率有待提升。因此本文提出新剪枝策略来优化DCI_Closed算法的搜索空间，并据此提出改进算法DCI_ESCS，其次利用DCI_ESCS算法对ABC库存分类问题进行优化。研究内容如下：

（1）本文将储存所有2-项集支持度信息的ESCS结构（Estimated Support Co-occurrence Structure）应用到经典闭频繁项集挖掘算法DCI_Closed上，提出针对2-项集的ESCS剪枝策略，最终得到改进的DCI_ESCS算法。并在SPMF公开资源库中connect、pumsb、chess、pumsb_star、accidents五个数据集上、不同最小支持度阈值下进行实验，对比分析算法改进前后的时间性能。实验结果表明，改进的DCI_ESCS算法在事务和项集较长、较稠密的数据集上表现良好，时间效率均有一定程度的提高。

（2）本文将DCI_ESCS算法应用到物资关联性ABC库存管理优化问题上。首先对原始数据进行初步分类作为后期分类调整的依据，然后用改进的DCI_ESCS算法挖掘出闭频繁模式并筛选出其中有效的模式，最后利用有效模式各物资的关联性调整部分初始分类。通过进一步分析可以得出，调整后的库存分类能够优化库存管理，降低库存成本，而且也能提高服务水平。

The reasonable classification of inventory materials is very important to the inventory management activities of enterprises. For each kind of materials, it can reduce the inventory cost and optimize the inventory management activities by formulating the corresponding inventory control strategy according to its nature. Among them, ABC inventory classification method is the most basic method. Most of the existing ABC classification studies only consider the independent demand for materials, without considering the internal correlation between materials. Frequent itemsets mining algorithm of frequent patterns contain the inner link between the material and can be used to optimize the ABC inventory classification, but a huge number of mining the frequent patterns can lead to some repeated, contradictions in the adjustment may occur when the inventory classification problem, and the use of the characteristics of frequent closed itemsets compression condition can solve this problem well. Closed frequent itemsets are a simplified representation of related frequent itemsets. Mining closed frequent itemsets greatly reduces the number of frequent itemsets in mining results, which has become an important research topic in data mining in recent years. Several closed frequent item set mining algorithms have been proposed, which can be applied to the ABC inventory classification optimization problem. DCI_Closed algorithm is a classic closed frequent item set mining algorithm. Through analysis, it is found that there is still room for improvement in pruning strategy and the efficiency of the algorithm needs to be improved. Therefore, this paper proposes a new pruning strategy to optimize the search space of DCI_Closed algorithm, and then proposes an improved algorithm DCI_ESCS. Then, the DCI_ESCS algorithm is used to optimize the ABC inventory classification problem. The research contents are as follows:

(1) In this paper, the Estimated Support co-occurrence Structure (ESCS), which stores all the Support information of 2-item sets, is applied to the classical closed frequent item set mining algorithm DCI_Closed, and an ESCS pruning strategy for 2-item sets is proposed. Finally, the improved DCI_ESCS algorithm is obtained. Experiments were carried out on five data sets of CONNECT, PUMSB, Chess, PUMSB_STAR and Accidents in SPMF open resource database under different minimum support thresholds to compare and analyze the time performance of the algorithm before and after improvement. Experimental results show that the improved DCI_ESCS algorithm performs well on long and dense data sets, and its time efficiency is improved to some extent.

(2) In this paper, DCI_ESCS algorithm is applied to the optimization problem of material-related ABC inventory management. First, the original data are preliminally classified as the basis for classification adjustment in the later stage. Then, the closed frequent patterns are mined and the effective patterns are screened by the improved DCI_ESCS algorithm. Finally, some initial classifications are adjusted by the correlation of the materials in the effective patterns. Through further analysis, it can be concluded that the adjusted inventory classification can optimize inventory management, reduce inventory cost, and improve service level.