A Synopsis Based Approach for Itemset Frequency Estimation over Massive Multi-Transaction Stream

The streams where multiple transactions are associated with the same key prevalent in practice, e.g., a customer has shopping records arriving at different time. Itemset frequency estimation on such is very challenging since sampling based methods, as popularly used reservoir sampling, cannot be used. In this article, we propose novel k -Minimum Value (KMV) synopsis method to estimate of itemsets over multi-transaction streams. First, extract KMV synopses for each item from stream. Then, estimator an itemset synopses. Comparing existing estimator, our not only more accurate and efficient calculate but also follows downward-closure property. These properties enable incorporation new frequent mining (FIM) algorithm (e.g., FP-Growth) mine To demonstrate this, implement FIM by integrating into algorithms, prove it capable guaranteeing accuracy bounded size synopsis. Experimental results massive show can significantly improve both estimating compared estimators.