Research On Computation Method Of Constraint Solving Complexity Based On Machine Learning

There may be many paths from the function entry to the target or function exit.Choosing a path with low constraint solving complexity can improve the efficiency of path-oriented test case generation,which is of great significance.Path-oriented constraint solving complexity can be reflected by the solution time.This paper first analyzes the key factors affecting the efficiency of path-oriented constraint solving.Then,a symbolic constraint solving complexity model is established by combining the composition of constraint information in the path.Finally,based on the constraint solving complexity model,error backpropagation neural network,error backpropagation neural network optimized by genetic algorithm,support vector machine regression and integrated learning are used to train a large amount of historic path solving data,and a predictive model is established to predict the solution time of a new set of input paths.Among them,the path with the least solution time is the path with the lowest solution complexity.This article encodes and implements the above four methods on MATLAB and verifies the feasibility of these methods through relevant experiments.The experimental results show that the error back propagation neural network model optimized by genetic algorithm has the highest prediction accuracy.Applying this model to the path selection of test case generation can improve the overall efficiency.