A Study on the Performance Enhancements of Association Rules Mining

• Main Authors: Huang, Sheng-Jhih, 黃聖智
• Other Authors: Chen, Chuei-cheng
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/80086714559121652754

碩士 === 南台科技大學 === 資訊管理系 === 99 === With the rapid development of information industry today, Major companies with the information technology and the amount of information is constantly increasing, So how can this huge database of information to make effective use of business in recent years has been the subject of frequent discussion. Therefore, data mining techniques are proposed and study its multi-faceted. In the data mining technology, association rule is the most widely used technology. As network technologies developed and global business model, business transaction databases can be said that continued non-stop expanding. In a large database, the traditional way of data mining will added the entire database into mining, Cause the larger database will have lower efficiency of mining. At the same time, it excessive redundant candidate projects let the database will scan too many times; It must be scanned the extra data to determine the location of the target data; and correctness of the results for the sake of exploration, the original data has been mining repeatedly. These are all traditional Apriori association rules mining algorithms shortcomings in the database in the implementation of mining. This study is based on Apriori algorithm, to improve its calculation process to make a new algorithm. The algorithm will transform the traditional standard database into the vertical database, and placed the minimum number of occurrences in the portfolio front, to reduce the combination and improve the efficiency; then add the concept of clustering, After all the data clustering, mining target cluster and cluster before target, to reduce redundant Data scan time. Finally, we use prediction. In the same time of scanning, screening the item which cannot be frequent itemset and remove it. It is no need to scan each data, as long as the scanning process found that cannot be the frequent itemsets, then stopped the scan and removed the item. Further reduce redundant scanning time, to improve the efficiency of mining. Then it based on above three methods to propose a new algorithm. This algorithm can maintain its accuracy among mining, and reduce the time and calculate memory of the mining, making it more consistent with enterprise requirements, to provide timely and correct decision-making.