Title

Using Time Series Models for Defect Prediction in Software Release Planning

Presenter Information

James Tunnell

Document Type

Oral Presentation

Location

SURC 137B

Start Date

21-5-2015

End Date

21-5-2015

Keywords

Software Defect Prediction, Release Planning, Time Series Model

Abstract

To produce a high-quality software release, sufficient time should be allowed for testing and fixing defects. Otherwise, there is a risk of slip in the development schedule and/or software quality. A time series model is used to predict the number of bugs created during development. The model depends on the previous numbers of bugs created. The model also depends, in an exogenous manner, on the previous numbers of new features resolved and improvements resolved. This model structure would allow hypothetical release plans to be compared by assessing their predicted impact on testing and defect-fixing time. The VARX time series model was selected as a reasonable approach. The accuracy of the model appeared low for a single dataset, but the error was found to be normally distributed.

Faculty Mentor(s)

John Anvik

Department/Program

Computer Science

Additional Mentoring Department

Computer Science

This document is currently not available here.

Share

COinS
 
May 21st, 8:50 AM May 21st, 9:10 AM

Using Time Series Models for Defect Prediction in Software Release Planning

SURC 137B

To produce a high-quality software release, sufficient time should be allowed for testing and fixing defects. Otherwise, there is a risk of slip in the development schedule and/or software quality. A time series model is used to predict the number of bugs created during development. The model depends on the previous numbers of bugs created. The model also depends, in an exogenous manner, on the previous numbers of new features resolved and improvements resolved. This model structure would allow hypothetical release plans to be compared by assessing their predicted impact on testing and defect-fixing time. The VARX time series model was selected as a reasonable approach. The accuracy of the model appeared low for a single dataset, but the error was found to be normally distributed.