Research On Software Quality Prediction And Evaluation Models And Methods Of Test Data Generation

The software industry is the fundamental, strategic and leading industry of the economic and social development. It plays an important role in promoting the convergence of information and industrialization, encouraging the industrial structure adjustment, transforming the mode of economic development and maintaining the state security. The research of software quality has promoted strategic and realistic significance in enhancing the quality of software products and in promoting the healthy and orderly development of the software industry. The modeling software quality prediction and evaluation and the method of automatic test data generation belong to the key technology of the related research of software quality, and become research focus in field of software engineering, which is provided with great theoretical significance and application value.The paper has set the study goals, research methods and technology roadmap based on the analyses of software quality in software industry and software quality-related technologies both domestically and abroad. Then the connotations and properties of software quality and software metric methods have been introduced, the typical software quality models have been discussed. This paper has carried out more in-depth research on the key technical issue related to software quality and discussed the key technologies in modeling software quality prediction and evaluation and the method of automatic test data generation based on intelligent optimization theory. The main contents of this paper are as follows:Firstly, On the basis of studying the principles and methods on generating test data, the method about automatic test data generation based on genetic-particle swarm hybrid algorithm has been proposed. The proposed method has replaced the style in genetic algorithm with the individual update method of particle swarm optimization, possessing both the merit of genetic algorithm on strong global searching and particle swarm optimization on strong local searching ability. The experimental results have shown that this method can keep the group diversity, avoid the premature convergence in evolution and have strong ability of global optimization and well efficiency of test data generation.Secondly, Prediction models are often used to find the nonlinear relationship between metric data and quality factors, on the basis of studying software quality prediction principles and methods, software quality prediction model base on PSO-BP network has been raised. It has indicated that the proposed model can predict software quality quickly and avoid the emergence of local best value and overcome the lack of parameter evaluated by the expertise, and can accurately reflect the relationship of the internal and external properties of software quality.Thirdly, on the basis of studying software quality evaluation principles and methods, the software quality evaluation model based on fuzzy regression with asymmetric triangular fuzzy numbers has been put forward, which has taken fully into account the impact of customer satisfaction. This model has investigated software quality evaluation by modeling the relationship between the customer requirements and software features so as to produce high-quality software.