Application of genetic algorithms to optimize solution of filtering and prediction problems in dynamic program testing systems

Abstract

Subject of research: probabilistic time models of testing created to form complex stochastic connections between individual test elements and developed to detect certain groups of the web applications program errors.
 The purpose of research: substantiate the possibility of using genetic algorithms in the process of solving probabilistic testing problems based on a multi-particle filter and evaluate their effectiveness. The study provides fundamental methods to improve the accuracy of the posterior distribution of probabilistic testing models and the total number of matched with evidences samples.
 Methods and objects of research: object of the research is to solve filtering and smoothing problems for a probabilistic test model based on a multi-particle filter. Methods and algorithms based on the Monte Carlo method are presented, allowing, in combination with genetic algorithms, to increase the accuracy of obtaining posterior estimates of samples. This approach allows you to narrow the range of samples, as well as increase their consistency. The formation of each next sample will be carried out taking into account the previous ones through the use of crossover and mutation operations.
 The main results of research: as a result, the validity of the proposed approaches to solving filtration and prediction problems in the process of implementing testing procedures based on multi-particle filter algorithms and genetic algorithms was proved. The given practical results prove the constructiveness and scientific validity of the proposed methods and algorithms for solving web applications testing problems.