Please use this identifier to cite or link to this item:
Full metadata record
DC FieldValueLanguage
dc.contributor.authorPuranik, S.
dc.contributor.authorDeshpande, P.
dc.contributor.authorChandrasekaran, K.
dc.identifier.citationProcedia Computer Science, 2016, Vol.93, , pp.924-930en_US
dc.description.abstractWith the growing complexities of the software, the number of potential bugs is also increasing rapidly. These bugs hinder the rapid software development cycle. Bugs, if left unresolved, might cause problems in the long run. Also, without any prior knowledge about the location and the number of bugs, managers may not be able to allocate resources in an efficient way. In order to overcome this problem, researchers have devised numerous bug prediction approaches so far. The problem with the existing models is that the researchers have not been able to arrive at an optimized set of metrics. So, in this paper, we make an attempt to select the minimal number of best performing metrics, thereby keeping the model both simple and accurate at the same time. Most of the bug prediction models use regression for prediction and since regression is a technique to best approximate the training data set, the approximations don't always fit well with the test data set. Keeping this in mind, we propose an algorithm to predict the bug proneness index using marginal R square values. Though regressions are performed as intermediary steps in this algorithm, the underlying logic is different in nature when compared with the models using regressions alone. � 2016 The Authors. Published by Elsevier B.V.en_US
dc.titleA Novel Machine Learning Approach for Bug Predictionen_US
dc.typeBook chapteren_US
Appears in Collections:2. Conference Papers

Files in This Item:
There are no files associated with this item.

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.