A Study on Data Mining from Aggregated Basket

2014 Fiscal Year Final Research Report

• Project/Area Number: 24500164
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Intelligent informatics
• Research Institution: The University of Electro-Communications
• Principal Investigator: NUMAO Masayuki 電気通信大学, 情報理工学(系)研究科, 教授 (90508821)
• Co-Investigator(Kenkyū-buntansha): MARUYAMA Hiroshi 統計数理研究所, モデリング研究系, 教授 (90609728)
• Project Period (FY): 2012-04-01 – 2015-03-31
• Keywords: 知識発見とデータマニング / バスケット分析 / ビッグデータ / 相関分析

Outline of Final Research Achievements

Basket analysis is a well-known data mining method to find the pattern of which item is purchased often with other items from point-of-sales (POS) data. In existing algorithm, however, the input basket is restricted to a set of items. Since the number of items in the set is 1 each at the most, it cannot deal with the aggregated basket where multiple baskets are merged into 1 basket. Thus it cannot deal with the basket of family items, or data collected in a certain period.
In this study, we defined the aggregated basket, proposed how to reconstruct the baskets from the aggregated basket to find the association pattern. We also developed the association mining system form the aggregated baskets and evaluated it by using the actual manufacturing data. The result shows that the proposed method is very useful to find the new patterns which cannot be extracted by the existing method, and it can be applied to manufacturing and distribution industry.

Free Research Field

知能情報学