Academic
Publications
Predicting build failures using social network analysis on developer communication

Predicting build failures using social network analysis on developer communication,10.1109/ICSE.2009.5070503,Timo Wolf,Adrian Schröter,Daniela Damian,

Predicting build failures using social network analysis on developer communication   (Citations: 23)
BibTex | RIS | RefWorks Download
A critical factor in work group coordination, communication has been studied extensively. Yet, we are missing objective evidence of the relationship between successful coordination outcome and communication structures. Using data from IBM's Jazztrade project, we study communication structures of development teams with high coordination needs. We conceptualize coordination outcome by the result of their code integration build processes (successful or failed) and study team communication structures with social network measures. Our results indicate that developer communication plays an important role in the quality of software integrations. Although we found that no individual measure could indicate whether a build will fail or succeed, we leveraged the combination of communication structure measures into a predictive model that indicates whether an integration will fail. When used for five project teams, our predictive model yielded recall values between 55% and 75%, and precision values between 50% to 76%.
Cumulative Annual
View Publication
The following links allow you to view full publications. These links are maintained by other sources not affiliated with Microsoft Academic Search.
    • ...Wolf et al. [31] used properties of social networks to predict the outcome of integrating the software parts within teams...

    Irwin Kwanet al. Does Socio-Technical Congruence Have an Effect on Software Build Succe...

    • ...There is a large volume of work on predicting failures at later stages of development: techniques have been proposed to predict run-time failures from source code metrics[16][23], to predict build failures from the communication structure between system developers (who communicated with who) [22], and to predict the system reliability from test case results [15]...
    • ...These failure types extend the usual product failures and integration problems that have been the focus of failure prediction techniques applied at later stage of the life-cycle [16][22] to include process failures such as abandoned development, stalled development and rejection reversal, which are specific to early failure predictions...
    • ...This means that, unlike other failure prediction methods that compare themselves against a random predictor [16][6][22], the baselines in our context are the most optimistic and most pessimistic predictors...
    • ...Some of these attributes could be more complex than the ones we have used here, such as for example an analysis of the communication structure between stakeholders [22], the roles of the users involved in a discussion, or the system components affected by a feature request...

    Camilo Fitzgeraldet al. Early failure prediction in feature request management systems

    • ...[9, 10]), the characteristics of the leaders and manager as well as their leadership styles [33], the patterns of interaction on task tracking systems [34] as well as the number of individuals involved in the development of a piece of software (e.g...

    Marcelo Cataldoet al. Factors leading to integration failures in global feature-oriented dev...

    • ...Other research links these metrics with performance statistics related to a success factor, supporting their use as performance indicators [6, 7]. Knowledge management research [8] suggests that both explicit (knowwhat) and tacit (know-how) knowledge capture take place for better shared information among software teams...
    • ...In RQ2, we found that WPM Scores performed the best as performance indicators, with a correlation (0.43 < r < 0.46), validating previous work by others [6, 7, 16] showing content quality and participation to be indicative of performance...
    • ...Wolf et al. [7] use social network analysis to suggest developers and managers can use communication metrics to predict final performance...
    • ...By measuring the levels of social connectedness among team members in project groups using IBM's Jazz [26], and correlating the quality of communication artifacts to their corresponding development artifacts they are able to predict failed software module integration of a subgroup‟s code with the larger software product [7]...
    • ...Thus, it is possible for managers to "adjust collaborative tools or processes towards improving the [quality of] integration [7]"...
    • ...Similar to other researches [7, 16], we are interested in how performance indicators might be combined to yield real time predictions of impending build or challenge shortcomings...

    Matthew Haleet al. Predicting individual performance in student project teams

    • ...Recent empirical case studies of several well-known software products have shown that SNA metrics are predictive of faults [11, 13], failures [10, 18], and vulnerabilities [8, 9, 16]...
    • ...The most prevalent of these three types of studies are the prediction studies [8-11, 13, 16, 18]...
    • ...Other case studies include the Jazz platform [18], Microsoft Windows [11, 13], a Nortel product [10], Mozilla Firefox [16], and the Linux kernel [8, 9, 16]...
    • ...Most researchers [4, 7-12, 16, 18] use logs from the version control system (e.g...
    • ...Some researchers [7, 18] apply communications logs (e.g...
    • ...Other studies [4,15,18] implicitly use an edge window of infinity, thus defining an edge where two developers change the same source code file at any time in history...

    Andrew Meneelyet al. Socio-technical developer networks: should we trust our measurements?

Sort by: