Font Size: a A A

The Investigation Of Metamorphic Relations And Their Key Features

Posted on:2020-07-07Degree:MasterType:Thesis
Country:ChinaCandidate:J W ZhangFull Text:PDF
GTID:2518305897970799Subject:Software engineering
Abstract/Summary:PDF Full Text Request
With the expansion of data in modern society,the application of data mining algorithms is more and more common.However,how to ensure the quality of programs in this field is currently a challenge.It is difficult to verify the correctness of the outputs of programs in data mining algorithms or the cost is too high,which makes it difficult for us to detect failures in this field by using traditional testing methods.That is the so called oracle problem.However,most of the research is focused on optimizing existing models or building more accurate models in data mining,while very little research has been done to ensure the correctness for algorithms themselves.Metamorphic testing(MT)is a simple yet effective testing technique to alleviate oracle problems.In metamorphic testing,new test cases are generated from existing test cases.The expected outputs of the new test cases can be checked by using necessary properties,so called metamorphic relations(MRs),among two or more input data and their expected outputs.We will investigate the metamorphic relations from two aspects.1.Investigating the construction of the metamorphic relations for data mining algorithms and investigating the performance of these metamorphic relations.We construct a series of metamorphic relations for data mining algorithms to test and verify the correctness of the implementation and the results returned.By conducting a series of empirical experiments,we demonstrate that our method has good performance on testing and verifying the correctness of the data mining algorithms.2.Mining the key features of the metamorphic relations.The resources for software testing are always limited,it requires us to select fewer and more efficient metamorphic relations to test the programs.Starting from the components of the metamorphic relations,we explore features that may affect the performance of metamorphic relations from the perspective of their input relations,output relations and the combination of metamorphic relations.By conducting a series of empirical experiments combined with cross-validation,we conclude a set of key features that may guide people to construct or select more effective metamorphic relations.
Keywords/Search Tags:Metamorphic Testing, Metamorphic Relations, Data Mining, Key Features of Metamorphic Relations
PDF Full Text Request
Related items