System Study Of Software Defect Mining With Weak Label

Software is an important part of the computer system.Software system is becoming more and more complex.Ensuring the quality of software systems has become a big challenge for researchers.The main cause of system unreliability is the software defect.There are a number of research work on software defect detection,which mostly studies the performance of a set of software measures or a specific defect detection algorithm,and usually requires a large number of labeled instances.In practical applications,software defect detection is usually weak labeled due to limitations of test resources.In this paper,we study and analyze the software defect mining system with weak label.Our contributions are as follows:1.Aiming at the problem that the weak label software defect mining has no suitable feature set,we propose to use information gain to select a proper feature set.This feature set can effectively represent the software module and provide the basis for software defect mining2.With no software detect information,we propose a method to find the module with high risk of defect by mining the exception module in the software.We compare the performance of common anomaly detection algorithms used in software defect mining and choose the proper one for our system.The initial feedback information can be given to the user without any defective information.Our system gives the user the initial feedback information when there is no label information3.With few software defect information,we propose active semi-supervised learning to detect defective modules.We compare the performance of active semi-supervised learning algorithm and traditional machine learning algorithm and we choose active semi-supervised learning algorithm for our system,which can reduce the number of software defects that users need to label.4.Based on the research above,this paper develops a software defect mining system with weak label,which works in different weak label conditions,and provides users with defect information of software module.