Özet:

Abstract The traditional frequent item-set mining is most popular and widely used technique for mining of related items. It considers whether the item is present or absence in dataset. However, item quantity and its importance is need to be consider for some real-world problem such as identify profitable items from the customer transaction dataset in supermarket, discover valuable customer for business, in medical field identify the combination of symptoms that are more significant to diseases. High utility itemset mining considers item quantity and its importance. Many researches have been done on the high utility itemset mining. Among them, utility list-based methods are efficient as it does not generate the candidate set. However, drawback of such techniques is lot of expensive join operations on utility list which degrades the performance of algorithm by increasing the storage requirement and time for execution. We proposed Predicted Utility Co-Exist Structure known as PUCS to store the utility data and Predicted Utility Co-Exist Pruning known as PUCP to eliminate unnecessary utility list join operations. It improves the algorithm’s performance. We experiment the proposed approach on standard real-life datasets and results are compared with existing methods. According to experimental result analysis, proposed PUCP-miner outperforms existing approaches concerning execution time and memory requirement. In terms of execution time, proposed approach achieves more than 20 % improvement and for memory consideration, proposed approach got 3% improvement compared to state of the art approaches.

Anahtar Kelimeler: