High Utility Mining of Streaming Itemsets in Data Streams

Abstract

The traditional models for mining frequent itemsets mainly focus on the frequency of the items listed in the respective dataset. However, market basket analysis and other domains generally prefer utility obtained from items regardless of their frequencies in the transactions. One of the main options of utility in these domains could be profit. Therefore, it is significant to extract items that generate more profit than items that occurs more frequently in the dataset. Thus, mining high utility itemset has emerged recently as a prominent research topic in the field of data mining. Many of the existing researches have been proposed for mining high utility itemset from static data. However, with the recent advanced technologies, streaming data has become a good source for data in many applications. Mining high utility itemset over data streams is a more challenging task because of the uncertainty in data streams, processing time, and many more. Although some works have been proposed for mining high utility itemset over data streams, many of these works require multiple database scans and they require long processing time. In respect to this, we proposed a single-pass fast-search model in which we introduced a utility factor known as utility stream level for tracing the utility value of itemsets from data streams. The simulation study shows that the performance of the proposed model is more significant compared with the contemporary method. The comparison has been performed based on metrics like process-completion time and utilized search space.