Research On Classification Algorithm Based On CAPE Over Data Streams

With the rapidly developing of Network, Information Technology and Database, A new form of data which is different from traditional data forms appeared, It's named of data streams by researchers. Data streams mining became a popular topic in recent years. Classification which played an important role in data mining, is continuously to be a significant part in data streams.Data streams has some new characteristics, data dose not take the form of persistent relations, but arrives in a rapid, multiple, continues, and time-varied way. The traditional classification methods such as Bayes, Decision-tree, Neural Network and so on, which used to be used in static data mining , can not work on data streams. This thesis uses classification algorithm based on frequent patterns. It means establishing classifier by mining frequent patterns from the first window, with the arriving of data streams, classifying data by using the classifier and updating the classifier at the same time. By studying and analyzing the algorithms of frequent pattern mining and classification. This thesis designs a new classification algorithm using frequent patterns over data streams based on CAPE algorithm.The new algorithm which used frequent items as classifier, and solved the problem of concept drift by using cache table, has a good improvement in memory usage and veracity in classification, what is proved by experiment, it also injects new idea and direction for the research of data streams mining.