Date of Award
CC BY-NC-ND 4.0
As new transactions update data sources and subsequently the data warehouse, the previously discovered association rules in the old database may no longer be interesting rules in the new database. Furthermore, some new interesting rules may appear in the new database. Generally, the process of generating new association rules using only the updated part of the database and the previously generated association rules is called incremental association rules maintenance. A straightforward approach for generating association rules in the new database starts from scratch. Obviously, this approach is not efficient and is time consuming since many computations could be repetitive work. In order to save some computation time and reduce maintenance cost, an incremental approach utilizes the previous association rules results to generate new association rules in the updated database. Although existing incremental approaches avoid many recomputations and improve the performance of maintaining incremental association rules, some overheads for getting association rules in the updated database still exist. This thesis proposes MAAP algorithm, an algorithm for maintaining incremental association rules aimed at making full use of the previous association rules results, reducing repetitive computations and which leads to better performance for incremental association rules maintenance. MAAP algorithm is based on the Apriori property and quickly generates some low level large itemsets from high level large itemsets, thus avoiding many repetitive computations. Meanwhile, MAAP algorithm stores delete part transactions and insert part transaction into an array respectively by scanning delete part database and insert part database one time. Thus it avoids many database scans. The developed method presents better performance over some existing algorithms and reduces maintenance cost in some situations. Some experiments for conducting comparative analysis between Apriori, FUP2 and MAAP algorithm have been included. Paper copy at Leddy Library: Theses & Major Papers - Basement, West Bldg. / Call Number: Thesis2000 .Z554. Source: Masters Abstracts International, Volume: 40-03, page: 0731. Adviser: Christie Ezeife. Thesis (M.Sc.)--University of Windsor (Canada), 2001.
Zhou, Zequn., "Maintaining incremental data mining association rules." (2001). Electronic Theses and Dissertations. 2660.