Mining Frequent Episodes on Data Streams

• Main Authors: Chia-Chang Nien, 粘嘉菖
• Other Authors: Show-Jan Yen
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/03264312408289986161

碩士 === 銘傳大學 === 資訊工程學系碩士班 === 97 === Data mining is a useful technique for data analysis, One of kind is mining frequent episodes in sequences. Users can predict event by mining frequent episodes in the future. They use level-wise in traditional methods, for example, candidate episodes are generated firstly, and we will find the frequent episodes by mining sequences. They spend a lot of time on scanning sequences and searching candidate episodes. In addition, data add with time in sequences continuously, it is called data streams. They consume time in scanning sequence. Data streams can be use in many applications, such as Intrusion Detection, web log, and meteorology. Users need to know the result immediately, but they must mining sequence by using previous methods again. For this reason, in this paper, we propose a method that they can only mining for additional data on data stream. They do not need to scan sequences and search candidate episodes, and only update frequent episodes.