![]() | Only 14 pages are availabe for public view |
Abstract Due to the development in technology, a number of applications such as smart mobile phone, sensor networks and GPS devices produce huge amount of ubiquitous data in the form of streams. Different from data in traditional static databases, ubiquitous data streams typically arrive continuously in high speed with huge amount, and changing data distribution. Dealing with and extracting useful information from that data is a real challenge. This raises new issues that need to be considered when developing association rules mining techniques for these data. It should be noted; that data, in the real world, are not represented in binary and numeric forms only, but it may be represented in quantitative values. Thus, using fuzzy sets will be very suitable to handle these values. In this thesis, the problem of mmmg fuzzy association rules from ubiquitous data streams is studied, and an efficient technique FFPUSTREAM (Fuzzy Frequent Pattern Ubiquitous Streams) is proposed to discover the complete set of recent fuzzy frequent patterns from a high-speed ubiquitous data streams over a sliding window model. In addition, a novel tree structure Dynamic Fuzzy Frequent Pattern tree (DFFP-tree) that combines fuzzy frequent pattern tree with the concept of dynamic tree restructuring at runtime was presented to efficiently mine fuzzy association rules Moreover, the complexity and the features of this technique were analyzed and discussed. Examples of real data sets are used to validate and test the proposed technique. Also, a medical case study is presented to demonstrate the usefulness and efficiency of this technique. Further research issues are also suggested. |