International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 11 Issue: 05 | May 2024
www.irjet.net
p-ISSN: 2395-0072
Analysis of Various Software Defect Prediction Techniques Pooja Gupta1, Komal Ahuja2 1M.Tech Student, Computer Science Engineering Department, GRIMT,Radaur(Yamunanagar), Haryana, India
2Assistant Professor, Computer Science Engineering Department, GRIMT,Radaur(Yamunanagar), Haryana, India
-------------------------------------------------------------------------***----------------------------------------------------------------------classifier using various machine learning techniques. Abstract
Previous research emphasises on enhancing the precision of predictions has mainly involved manually crafting discriminative features or combining characteristics [3]. Halstead features based on operators and operands, McCabe features based on dependencies, and CK features for object-oriented programs are some of the examples. However, traditional hand-crafted features often overlook the intricate semantics and clearly-defined syntax concealed in the Abstract Syntax Trees (ASTs) of programs.
Software defect along with an intrinsic element of software product is also an important aspect of software quality. Software defects are an unavoidable co product of the developed software. In addition to this, the guarantee of software quality assurance is not so easy and requires a lot of time too. There are different ways to define defects, such as in terms of quality. The defects may be present in all products whether it is a small program or large-scale software system. A number of defects can be discovered without any complexity. The various steps are applied for the software defect prediction which includes pre-processing, feature extraction and classification. The various techniques which are proposed in the previous time for the software defect prediction are reviewed and analysed in this paper. The techniques are reviewed in terms of technique description and their findings in terms of various parameters
Prediction models, important to Software Defect Prediction (SDP), are crucial in anticipating software faults. Number of methods and algorithms have been used to increase the correction of Software Defect Prediction (SDP) models, till the fundamental stages of SDP can be succinctly outlined. Primarily, collection of data includes collecting both pure and defective code samples from software repositories [4]. Next to this, a dataset is constructed by extracting related qualities from the compiled samples, with forthcoming steps including dataset balancing if any imbalance is identified, model training utilizing the prepared dataset, and the upcoming predictions of problematic components within new software datasets using the trained model. Consequently, the performance of the SDP model is evaluated. This innovative process allows for ongoing refinement and improvement over time.
Keywords - Software defect, Prediction, Machine learning, Pre-processing, Feature extraction, Classification
1. Introduction Because of the increasing complexity of today’s software and enhance chances of failures, ensuring reliability has become an important concern. Organisations such as Google employ code review and unit testing to identify problems in new code and increases reliability. Meanwhile, testing each code unit is not practical and human code reviews are labour- intensive [1]. With less funding for software projects, it is fruitful to detect potential issues immediately. As a result, software defect prediction algorithms are commonly employed to automatically identify possible mistakes, enabling developers to make optimal consumption of their resources. Software defect prediction includes, creating classifiers that analyse data such as change history and complexity of code to identify code segments with potential flaws. This practice ensures code reviewers to allocate their efforts with more strategy and receive warnings about potentially buggy code regions based on the prediction outcomes [2]. These code sections may include alterations, files, or processes. In the typical fault prediction process, there are two main stages: feature extraction from source files and the creation of a
© 2024, IRJET
|
Impact Factor value: 8.226
Collect clean & defective code samples
Extract features
balance dataset (optional)
Train SDP model
Predict defects
Evaluate the performance of SDP model
Figure 1: Software Defect Prediction Process
|
ISO 9001:2008 Certified Journal
|
Page 2235