Change Itemset Mining in Data Streams
Data stream mining is becoming very important in many application areas, such as the stock market. A data stream consists of an ordered sequence of instances and because there are usually a large number of instances along with limited computing and storage capabilities, algorithms that read the data only once are preferred. There has been some research that focuses on finding when a concept has changed, given some knowledge about the previous instances in the data stream, but little on determining the characteristics of that change. In this paper we concentrate on finding the characteristics of the changes that occur, using frequent itemset mining techniques. We propose an approach, which combines both heuristic and statistical approaches, that analyses changes that have occurred within a stream at itemset level and identify three types of change: extension, reduction, and support fluctuation. We evaluate our algorithm using both synthetic and real world datasets.
KeywordsChange Mining Data Stream Frequent Itemset Hoeffding Bound
Unable to display preview. Download preview PDF.
- 2.Agrawal, R., Psaila, G.: Active data mining. In: Fayyad, U.M., Uthurusamy, R. (eds.) KDD, pp. 3–8. AAAI Press (1995)Google Scholar
- 4.Steinbrecher, M., Kruse, R.: Identifying temporal trajectories of association rules with fuzzy descriptions. In: Annual Meeting of the North American Fuzzy Information Processing Society, NAFIPS 2008, pp. 1–6 (2008)Google Scholar
- 5.Liu, B., Ma, Y., Lee, R.: Analyzing the interestingness of association rules from the temporal dimension. In: Proceedings of the 2001 IEEE International Conference on Data Mining. ICDM 2001, pp. 377–384. IEEE Computer Society, Washington, DC (2001)Google Scholar
- 11.Leung, C.S., Khan, Q.: Dstree: A tree structure for the mining of frequent sets from data streams. In: Sixth International Conference on Data Mining, ICDM 2006, pp. 928–932 (2006)Google Scholar