SINGLE-PASS FAST-SEARCH HIGH UTILITY STREAMING ITEMSET MINING IN DATA STREAMS

Abstract

The conventional methods of itemset mining rely on the occurrence of the items listed in the corresponding itemset.  Nevertheless, the domains such as market basket prefer the utility that gained from an item instead of its occurrence in the transactions. Mandate option of the utility in such domains is profit. Hence, it is obvious to notice the items, which produces more profit than the items that occur more in the transaction. Hence, utility-based itemset mining is the buzz of current demining practices. Most of the contemporary contributions of utility-based itemset mining are limited to their stationary scope data. The buzz of the data sources in the recent past is data streams, which continuously produces the data, which may or may not contain labels for the elements involved in the record formation. The constraints, such as lack of knowledge on the records, which streams in the next moment enlighten the search space, process time as crucial compared to stationary data. In this regard, the proposal of this manuscript is a high utility itemset mining that introduced a utility parameter called stream level utility and a single pass fast search technique to track the itemsets of the streaming data. The experimental study denoting the significance of the proposal compared to the other contemporary model — the performance estimation carried by using the metrics called search space used, process completion time.