A Request-Based Frequent Pattern Mining Algorithm in Multi-Store Environment

• Main Authors: PENG, CHI-YUAN, 彭琪媛
• Other Authors: HU,HSIAO-WEI
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/25p78j

碩士 === 東吳大學 === 巨量資料管理學院碩士學位學程 === 107 === In today’s business environment, most of the companies have branches, subsidiaries, chain stores or dealers. And these branches could be located in many different geographical locations. However, the traditional frequent pattern mining algorithm might not be suitable in the multi-store environment nowadays. Therefore, over the last few decades, there have been many studies focus on advanced frequent pattern mining algorithm in a multi-store environment. However, we found that some of the existing methods consider only the multi-store environment problem but might not think about the cloud computing issues, for example, the cost and privacy issues in the cloud computing environment. In contrast, some of them overcome the cloud computing environment issue but not focus on frequent pattern mining in a multi-store environment. In this paper, we propose a new algorithm to remedy this research gap – A Request-Based Frequent Pattern Mining Algorithm (FFPM). We focus on the request-based algorithm and also expect that those frequent patterns under different stores and time periods are considered while considering those cost limitation and privacy problem in the cloud computing environment. To evaluate the efficiency and effectiveness of the proposed method, four experiments are performing in this study. The first experiment is about the running time of FFPM, and the second and third experiments are efficiency and effectiveness of FFPM compared with Apriori algorithm. The last experiment is evaluating FFPM by using a real data set, called SC-POS. Results of this study show that not only on the running time or precision but the proposed method can also find out those frequent patterns under different stores and time periods even more efficiently compared with the traditional frequent pattern mining algorithm.