Study Of Closed Iterative Sequential Patterns Mining Algorithm And Its Application In Software Dependability Analysis

Iterative sequential pattern mining is one branch of sequential patttern mining which is important role in data mining. And iterative pattern mining is widely used in practical life such as analysis of gene, market data and so on. In this thesis, we take an in-depth study on sequential pattern and iterative sequential pattern mining. And a new algorithm is proposed based on those studies. The main research contents and contributions include:Firstly, based on the analysis of the limitation belonging to the tradition iterative closed pattern mining, the closed iterative sequential pattern mining algorithm called CIPMPBE (Closed Iterative Pattern Miner via Prime-Block Encoding) is proposed. Based on sequence ID in the sequence database, all the sequence and a single sequence is encoded by prime-block. And then CIPMPBE mine the closed iterative pattern on the basis of prime-block as unit. So prime-block encoding save mining time and improve time efficiency of mining algorithm. The result of experiment prove the correctness and usefulness of the algorithmSecondly, forced on two sub-problems in the software dependability, which are the analysis of software behaviors failure and software specification discovery, CIPMPBE is used to solve those two problems. And two area of study is accomplished. Firstly, based on the importance of software specification, CIPMPBE is used to discover the software specification. CIPMPBE obtain a lot of order of software module invocation by running a lot of test case. And then CIPMPBE is used to find the closed iterative sequential pattern in order to discover the software specification. The analysis of sub-module invocation of JBOSS APP.SERVER is used to provide practical instance of the method. Secondly, based on analysis of software execution trace, CIPMPBE is used to construct classifier to analyze failure of software behavior. The method composes three steps. Firstly, the database containing great many execution trace of software transfer to the sequence database. Secondly, CIPMPBE is used to mine the sequence database. Thirdly, the closed iterative unique pattern mined by CIPMPBE is used to construct the classifier which can classify software execution trace for fault detection. The analysis of software behavior failure in Concurrent Versions System and Siemens Test Suite is used to provide practical instances of the method...