Incremental Association Rule Mining with Frequent Edge Graph and Dynamic Rule Generation for e-Commerce

Abstract

Association rule mining is one of the most common data mining tasks used to identify relations between items, which often appear in the same transaction data. This information can be used in such areas as boosting sales through promotional tactics that encourage purchases of items usually bought together. However, in today’s shops, more are turning to the e-Commerce platform to sell their products. As in a real shop, product numbers increase with new items being added, while a few are out of stock. These kinds of situations occur around the clock and for this reason, the current association rule mining algorithms are no longer suited for such dynamic. Therefore, a new algorithm called iARMFEG (incremental Association Rule Mining with Frequency Edge Graph) is proposed. The iARMFEG brings in information from each transaction one at a time to build an incremental weight graph. The resulting rules are generated from the constructed weight graph. An advantage of the proposed method is the ability to retaliate rule generation multiple times without having to go through the Frequency Itemset Generation process again. Moreover, it can also search for rules from each list of items or only from a specified. By doing so, the rare item problem is then resolved. The proposed method can successfully generate graphs by only reading data one time from a database, while FP-Growth must read two times. In addition, it was also found that the proposed method has a 95 percent faster processing time than Apriori methods because they also need to read data many times.