Algorithm Research Of Software Behavior Pattern From Software Execution Trace

With the development of global intelligence,computer software plays an increasingly significant role in our lives.We product many software products in various industries,more and more software data are appeared.The method of data mining is applied to the software behavior analysis,and the valuable information can be obtained from these software data,which can help the software developer to make the system error location,software vulnerability prediction,behavior prediction and so on,so as to effectively accomplish the work of software update and maintenance.In this paper,we provide information for dynamic analysis of software behavior.We mine sequences from the software execution trace,and provide the high utility sequences pattern for the researchers.The main work of this paper is as follows.First of all,we analyze the characteristics of software execution trace.Based on the theoretical study of software execution trace,this paper puts forward that the research work of this paper is based on the analysis of software dynamic behavior.We elaborate the method of obtaining software execution trace and analyze the characteristics of sequence pattern.At the same time,we summarize the ideas of classical algorithms in sequences pattern mining.We analyze the advantages and disadvantages of these algorithms.Secondly,an efficient closed pattern mining algorithm based on the execution trace of the software is proposed to solve the problem that the number of behavioral model is too big.A structure is designed for each function pattern in the software execution trace,preserving its sequence and utility values.According to this structure,we propose a pruning strategy,which can effectively reduce the number of expansions.At the same time,according to this structure,we design a calculation method to achieve the purpose of compressing the number of sets.Third,an algorithm for mining Top-k high utility continuous sequences pattern is proposed to solve the problem that the user can not determine a suitable utility threshold.We design a structure for each function in the software execution trace,which contains the next adjacent function information to exploit the continuous sequences pattern.At thesame time,four pruning strategies are proposed to discard the patterns after k positions as soon as possible,and the efficiency of the algorithm is improved in different aspects.Finally,we do experiments on the different datasets in the Windows environment.The effects of the proposed algorithms and other algorithms of the same type are evaluated from the aspects of running time,memory usages and the efficiency of strategies.