Frequent pattern mining over web data streams using sliding window model

Today, searching repetitive patterns on data flows is very important. By data flow we mean a type of data which is constantly produced in a very fast and unlimited manner. As a kind of these data we can name the report of clicks in computer networks. A repetitive pattern is a pattern which is available in a significant number of transactions. Finding repetitive patterns in data flows is a new and arguable issue in data mining as data is received in form of fast and continuous flow. Unlike static databases, flow mining faces a lot of problems including single review, requiring unlimited memory and high rate of input data. A common way of searching repetitive patterns is the excess check of data which requires to be saved in memory. In addition, according to the features of data flows i.e. unlimited and fast production, it is not possible to save them in memory and hence techniques are needed which are able to process them online and find repetitive patterns. One of the most popular relative techniques is using sliding windows. It sadvantage is reduction of the consumed memory and increasein search speed.In this paper, a new vertical display and an algorithm based on pins, called DBP-BA, are proposed to find repetitive patters in data flows. Since this new display without any additional task has a compact form, the proposed algorithm has a better performance than similar ones in terms of consumed memory and processing time. On the other hand, experiments support this matter.

کلیدواژه ها:

data flows ، sliding windows ، pin ، repetitive data pen set