Mining Method Based On Fault Effect On Behavior Pattern From Software Execution Graph

With the continuous expansion of software applications,the scale of software systems has become increasingly complex,consuming a great deal of time and effort from scientific and technical personnel.Excavating representative information from massive dynamic execution trajectories of software provides a complete basis for analyzing the structural features of software systems and software maintenance.This topic analyzes the software structure from the perspective of software fault propagation,combines data mining theory knowledge,and mines a representative behavior pattern of the software,thereby helping technicians to effectively complete the software update and maintenance work.Firstly,a software network model construction method based on software dynamic execution trajectory is proposed.By tracking the software execution trajectory,the function is used as a node,the function call relationship is used as the edge of the connection node,and the degree of dependency between the functions is used as the weight of the edge.It is mapped as a directed weighted software execution graph,and the software is traversed through the depth to obtain the software.Execution of the set of sequences provides the basis for mining behavior patterns.Secondly,an algorithm based on fault propagation to mine key nodes of software network is proposed.The algorithm considers the fault vulnerability of the function and the influence of the fault propagation.It defines the functions of the software system that have a high probability of fault and the range of the fault after the fault is a key node.The algorithm TPFPN-Miner is used to calculate and rank the fault effect.Force metrics,mining key function nodes in the network to improve the efficiency of fault location of complex software.Thirdly,this paper proposes an algorithm based on call relation to mine important software behavior patterns.This algorithm is based on the metrics of fault influence proposed in the previous section,and it is used as the external utility value of the function node in the software execution sequence.Combined with the continuity and sequence of software execution sequence,the algorithm FITK-Miner is designed and three strategies are proposed.Mining Top-k efficient behavior patterns in software sequences,and define it as an important behavior pattern of software based on the influence of function failure.Finally,the algorithm is tested on different data sets,the evaluation and analysis are carried out from different angles,the effectiveness of the proposed algorithm and the other same type algorithm is compared to verify the validity and correctness of the proposed method.