<![CDATA[As one of data mining applications, market basket analysis is generally performed using Apriori method. But, this method is tends to searching the association degree of items by only counting how many times the items appears on the overall transaction without concerning the items quantity on each transaction. Therefore, we propose a new concept which is based on perception that the more items bought in a transaction, the weaker the relation between items on that transaction. For the experimental purpose of the realization the concept, we collect a month sales transaction data from a supermarket. The data are transformed to another form that can be used by the software. Then, an algorithm is created to process that data in order to generate association rule of items in transaction. The rules can be previewed as table or graphic. Borland Deplhi 7 and MS Access 2003 is used in to build this experiment software. By using the output of this software, association rules, association degree of items, which is useful to help the decision maker to decide market policies. Based on the testing result, can be concluded that the smaller minimum support and confidence, the more rule to be generated and the more processing time needed. Also, the higher the combination count to be searched, the less processing time needed.]]>
Copyrights © 2012
