Research On Mining Algorithms For Software Behavior Path Based On Complex Network

Key paths and dissimilar paths in complex software executing network caneffectively analyze the executing behavior rules of software. Key paths and dissimilarpaths play an important role in software executing process and maintaining networkconnectivity. They have important significance for normal software executing processand related conclusions and methods can help to maintain software executing processbetter and improve software quality.Firstly, an approach called Graph-Mine is proposed to mine key paths in complexsoftware executing network. Complex software executing network graph (SENG) isdefined, and functions and call relationship among them are mapped to nodes andpaths respectively. DFS-SENG algorithm is adopted to transform complex softwareexecuting network graph (SENG) into software executing path (SEP). An adjacencyextension strategy is designed to avoid inappropriate mined results and improvemining efficiency. Key-Path-Search algorithm is presented to mine key nodes and keypaths from software executing paths.Secondly, a key path mining algorithm called Graph-Cluster-Mine in complexsoftware executing network modules is proposed. GN algorithm is adopted to clusterthe network into modules on limiting scale of a module. Then, DFS algorithm isadopted for each module to find software executing paths (SEP) in each module.Path-Simplified-Strategy-Bundle and Path-Simplified-Strategy-Hierarchy are devisedto simplify paths. Key-Path-Mine algorithm is developed to mine weighted key pathswithin the module. Weighted closed key paths of software executing are found byPath-Closed-Checking-Strategy.Finally, a dissimilar path detecting algorithm called Graph-Dissimilar-Detect incomplex software behavior network based on time-interval weight and dissimilarity isput forward. DFS algorithm is adopted for complex software behavior network toobtain software behavior paths. Path-Length-Constraint strategy is designed to prunesome useless paths. Time-interval-Constraint strategy is devised to prune some weak relation between software behaviors. KPath-Mine algorithm is designed to mineweighted key paths. Strong-Path-Checking strategy is developed to prune weightedkey sub-paths early to improve efficiency. Dissimilar-Path-Mine algorithm is devisedto find dissimilar paths with path dissimilarity.The experiments are conducted on NetBeans in java. The time cost, scalabilityand the number of key paths are considered to validate the effeciency of the proposedalgorithms.