Literature Survey and Scope of the Present Work

Literature Survey and Scope of the Present Work

DOI: 10.4018/978-1-5225-3185-2.ch002
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

As I know large numbers of techniques and models have already been worked out in the area of error estimation. Identifying and locating errors in software projects is a complicated job. Particularly, when project sizes grow. This chapter enlists and reviews existing work to predict the quality of the software using various machine learning techniques. In this chapter key finding from prior studies in the field of software fault prediction has been discussed. Various advantages and disadvantages of the methods used for software quality prediction, have been explained in a detail. What are the problems solved are also mentioned in this section. Description of earlier research work and present research work has summarized in one place.
Chapter Preview
Top

Literature Survey

Zhang and Tsai (2002) mention that machine learning methods are used to predict or estimate software quality, software size, software development cost, project or software effort etc. Machine learning algorithms have proven to be of great practical value in a variety of applications. The field of software engineering turns out to be a fertile ground where many software development and maintenance tasks could be formulated as learning problems and approached in terms of learning algorithms. In this work, they have discussed issues and current status regarding machine learning applications to software engineering problems. They considered machine learning methods that can be used to complement existing software tools. The strength of machine learning methods lies in the fact that they have sound mathematical and logical justifications and can be used to create and compile verifiable knowledge about the design and development of software artifacts (Zhang & Tsai, 2002). Mertik et al. (2006) presented the use of advanced tool for data mining called multimethod on the case of building software fault prediction model. Current software quality estimation models often involve use of data mining and machine learning techniques for building a software fault prediction model. In this work, they have presented some of the methods integrated within the multimethod data mining tool. They have introduced the multimethod data mining tool which was developed in the laboratory for the system design in University of Maribor (Lenic, n.d.), and it was presented as a case study of building the fault prediction model based on the data from the metrics data program data repository (Lenic, n.d.). In their study, they adapted and combined some single methods, approaches with the multimethod tool on the real data sets from the MDP data repository, where they got promising results. They have given an overview of some efficient single methods approach for support of multi-methods. These are decision tree (DT), support vector machine (SVM) and genetic algorithm (GA). With multi-method tools, they have generated four different fault prediction models. Each model has been built with the different techniques. In every experiment, they sampled the accuracy on learning set and test as also the size of the generated model/classifier. In this work, they have presented the advanced data mining tool multimethod for building software fault prediction models. They showed the use of the tool on three different datasets of the NASA IV & V Metrics data program project and using multimethod tool they got quite better results as with standard supervised machine learning methods for building such prediction models. Therefore, the reasons for which multimethod approach was appropriate for building the software fault prediction models are:

  • 1.

    It provides reasoning based knowledge in the form of a multimethod tree.

  • 2.

    It combines different single methods for building fault prediction software model.

Complete Chapter List

Search this Book:
Reset