TDMCS: An Efficient Method for Mining
Closed Frequent Patterns over Data Streams Based on Time Decay Model
Meng Han,
Jian Ding, and Juan Li
School
of Computer Science and Engineering,
North Minzu University, China
Abstract: In some data stream applications, the
information embedded in the data arriving in the new recent time period is
important than historical transactions. Because data stream is changing over
time, concept drift problem may appear in data stream mining. Frequent pattern
mining methods always generate useless and redundant patterns. In order to
obtain the result set of lossless compression, closed pattern is needed. A
novel method for efficiently mining closed frequent patterns on data stream is
proposed in this paper. The main works includes: distinguished importance of
recent transactions from historical transactions based on time decay model and
sliding window model; designed the frame minimum support count-maximal support
error rate-decay factor (θ-ε-f) to avoid concept drift; used closure operator
to improve the efficiency of algorithm; design a novel way to set decay factor:
average-decay-factor faverage in order to balance the high recall
and high precision of algorithm. The performance of proposed method is
evaluated via experiments, and the results show that the proposed method is
efficient and steady-state. It applies to mine data streams with high density
and long patterns. It is suitable for different size sliding windows, and it is
also superior to other analogous algorithms.
Keywords: data stream mining, frequent pattern mining, closed pattern
mining, time decay model, sliding window, concept drift.
Received January 15, 2015; accepted August 12,
2015