AT&T Labs Research - Bell, Robert M.

Robert Bell received a Ph.D. in Statistics from Stanford University. He has been a member of the Statistics Research Department at AT&T Labs-Research since 1998. He previously worked for 18 years at RAND doing public policy analysis. His current research interests include machine learning methods, analysis of data from complex samples, and record linkage methods. He was a member of an international team that won the Netflix Prize competition. He is a fellow of the American Statistical Association. Dr. Bell is a recognized expert on use of statistical methods in the decennial census, having served on four National Research Council panels advising the U.S. Census Bureau, one as chair, and the Census Advisory Committee of the American Statistical Association. He has served on the Fellows Committee of the American Statistical Association, the board of the National Institute of Statistical Sciences, and the Committee on National Statistics.

Team BellKor Talks About Netflix Prize

{Several studies have examined code churn as a variable for predicting faults in large software systems. High churn is usually associated with more faults appearing in code that has been changed frequently. We investigate the extent to which faults can be predicted by the degree of churn alone, whether other code characteristics occur together with churn, and which combinations of churn and other characteristics provide the best predictions. We also investigate different types of churn, including both additions to and deletions from code, as well as overall amount of change to code. We have mined the version control database of a large software system to collect churn and other software measures from 18 successive releases of the system. We examine the frequency of faults plotted against various code characteristics, and evaluate a diverse set of prediction models based on many different combinations of independent variables, including both absolute and relative churn. Churn measures based on counts of lines added, deleted, and modified are very effective for fault prediction. Individually, counts of adds and modifications outperform counts of deletes, while the sum of all three counts was most effective. However, these counts did not improve prediction accuracy relative to a model that included a simple count of the number of times that a file had been changed in the prior release. Including a measure of change in the prior release is an essential component of our fault prediction method. Various measures seem to work roughly equivalently. }

This version of the work is reprinted here with permission of IEEE for your personal use. Not for redistribution. The definitive version was published in Testing: Academic & Industrial Conference (TAIC 2011) , 2011-03-25

{Does the use of fault prediction models to help focus software testing resources and other development efforts to improve software reliability lead to discovery of different faults in the next release, or simply an improved process for finding the same faults that would be found if the models were not used? In this short paper, we describe the challenges involved in estimating effects for this sort of intervention and discuss ways to empirically answer that question and ways to assess any changes, if present. We present several experimental design options and discuss the pros and cons of each.}

(c) ACM, 2010. This is the author's version of the work. It is posted here by permission of ACM for your personal use. Not for redistribution. The definitive version was published in The 6th International Conference on Predictive Models in Software Engineering , 2010-09-12

{We investigate whether files in a large system that are modified by an individual developer consistently contain either more or fewer faults than the average of all files in the system. The goal of the investigation is to determine whether information about which particular developer modified a file is able to improve defect predictions. We also continue an earlier study to evaluate the use of counts of the number of developers who modified a file as predictors of the file's future faultiness. The results from this preliminary study indicate that adding information to a model about which particular developer modified a file is not likely to improve defect predictions. The study is limited to a single large system, and its results may not hold more widely. The bug ratio is only one way of measuring the 'fault-proneness' of an individual programmer's coding, and we intend to investigate other ways of evaluating bug introduction by individuals. }

Awards
Science & Technology Medal, 2003. Honored for technical leadership toward fair implementation of the Telecommunications Act.

AT&T Labs Research — Leading Invention, Driving Innovation

site map