Research And Implementation On Software Testing With Statistical Method

Software testing is an important stage of software engineering, it's also an important instrument to ensure software's quality.Software testing costs so much resource, but it lacks of great theory base. The testcases are always chosen by testers' experience, and also full of individual idea and subjectivity. So it results in that the software's quality isn't according to users' standard, and it reduces users' degree of satisfaction about the software's quality.This paper proposes a software testing method based on statistical. It is a method that can generate sufficient testcases, and it is also a method more fit for software practice. This method bases on the users' using mode, and pays more attention to frequent operation, so it's effective to find the mistakes that own much more influence. Statistical testing constructs software's Markov chain usage model firstly, generates testcases according to usage model, analyses the test result according to statistical model after test the software, then evaluate the software's reliability and quality. Rule of judging adequate testcase about software's reliability is deficient, so this paper proposes a new method that can generate sufficient testcases.With the basis of statistical testing theory, this dissertation designs and develops a statistical testing tool "MBCGA" based on Markov chain, it supports Markov chain usage model building, parameters analysis, test cases generation and analysis and so on. It helps testers to get testcases own strong sufficiency. It provides to testers an applied, easy-manipulative test tool based on statistical method.Using statistical testing tool "MBCGA" to test Vehicle engine compartment computer aided layout design system can validate software testing based on statistical method is a test method with strong availability. And it's a successful software test based on model, so it can evaluate and forecast the software's reliability.