Finding recently frequent itemsets adaptively over online transactional data streams,

Mining frequent patterns in a varying-size sliding window of online transactional data streams

“…Various techniques have been proposed for mining frequent patterns in online data streams [5][6][7]10,11,[13][14][15][16][19][20][21][22][23][24][25][26]. The four main design elements that characterize these techniques, as well as the proposed approach, are discussed below.…”

“…In order to differentiate the degree of importance of the patterns in recent transactions from that of those in historical ones, Chang and Lee [11,12] introduced a decay scheme [13] for stream data mining and developed an algorithm for maintaining frequent patterns over data streams, with each transaction assigned a weight that constantly changes according to its age in the window. In this way, older transactions contribute less towards pattern frequencies, allowing recent significant patterns in the data stream to be easily selected.…”

“…In this way, older transactions contribute less towards pattern frequencies, allowing recent significant patterns in the data stream to be easily selected. The time-fading model of Chang and Lee [11,12] is a variation of the landmark model [7], which cannot be adapted to mine frequent patterns in a sliding window, not to mention a varying-size window.…”

Frequency-based load shedding over a data stream of tuples

“…Therefore, even though there is no overload in the data stream, an unnecessary load shedding operation is performed in these load shedding techniques. However, considering that the rate of a data stream is more likely to be changed over time [9,12,13,26], the number of data elements to be alive should be dynamically controlled by the changing rate of the data stream as well as the capacity of its main-processing operation. Tatbul et al [23] proposed an adaptive load shedding technique over data streams in the Aurora system, which can be performed regionally or be stopped by considering the loads in the data streams.…”

Kalman Filter based Flexible Sliding Window Algorithm for Mining Frequent Itemset over Data Stream

“…This change is the results of a change in the parameters of the model [11]. The concept change makes frequent itemset mining in data streams even more challenging than traditional static databases.…”