Utility-based Frequent Itemsets in Data Streams using Sliding Window

There is a need for increased use of Market Basket Analysis in Economics to estimate consumer behaviour and demand function more realistic especially in a data streaming environment, which is a challenging task. A sliding window contains the latest fixed number of elements of the data stream. The algorithm FIMIU, proposed in this paper, replaces the itemsets in the sliding window by pointers to a single copy of the itemset, thereby creating more space for new itemsets in it which allows the user to analyze a bigger part of the data stream at a time. Experiments have shown that the proposed algorithm is memory efficient, however requires a bit extra time.