Skip to Main Content
Market basket analysis is very important to everyday's business decision, because it seeks to find relationships between purchased items. Undoubtedly, these techniques can extract customer's purchase rules by discovering what items they are buying frequently and together. Therefore, to raise the probability of purchasing the corporate manager of a shop can place the associated items at the neighboring shelf. For these reasons, the ability to predict e-shopper's purchase rules basing on data mining has become a competitive advantage for the company. On the other hand, mining maximal frequent patterns are also a key issue to the recent market analysis since; a maximal frequent pattern for a particular customer reveals the purchase rules. Moreover, if the dataset is sparse due to the presence of null transactions, the mining performance degrades drastically in existing approaches. In this paper, first we remove null transactions from the original dataset then we apply the bottom-up row enumeration tree approach to generate the maximal frequent patterns; later on the modified version of the sequence close level is used for counting the distance between a pair of items for mining the customer's purchase rules in an online transactional database. Experimental results show that our proposed approach is superior to previous approaches and can predict more accurate customer's purchase rules in reasonable time.