Rights statement: © ACM, 2016. 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 ACM Transactions on Knowledge Discovery from Data, 10, 3, 2016 http://doi.acm.org/10.1145/2798730
Accepted author manuscript, 1.79 MB, PDF document
Available under license: CC BY: Creative Commons Attribution 4.0 International License
Final published version
Research output: Contribution to Journal/Magazine › Journal article › peer-review
Research output: Contribution to Journal/Magazine › Journal article › peer-review
}
TY - JOUR
T1 - Mining user development signals for online community churner detection
AU - Rowe, Matthew
N1 - © ACM, 2016. 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 ACM Transactions on Knowledge Discovery from Data, 10, 3, 2016 http://doi.acm.org/10.1145/2798730
PY - 2016/2
Y1 - 2016/2
N2 - Churners are users who stop using a given service after previously signing up. In the domain of telecommunications and video games, churners represent a loss of revenue as a user leaving indicates that they will no longer pay for the service. In the context of online community platforms (e.g., community message boards, social networking sites, question--answering systems, etc.), the churning of a user can represent different kinds of loss: of social capital, of expertise, or of a vibrant individual who is a mediator for interaction and communication. Detecting which users are likely to churn from online communities, therefore, enables community managers to offer incentives to entice those users back; as retention is less expensive than re-signing users up. In this article, we tackle the task of detecting churners on four online community platforms by mining user development signals. These signals explain how users have evolved along different dimensions (i.e., social and lexical) relative to their prior behaviour and the community in which they have interacted. We present a linear model, based upon elastic-net regularisation, that uses extracted features from the signals to detect churners. Our evaluation of this model against several state of the art baselines, including our own prior work, empirically demonstrates the superior performance that this approach achieves for several experimental settings. This article presents a novel approach to churn prediction that takes a different route from existing approaches that are based on measuring static social network properties of users (e.g., centrality, in-degree, etc.).
AB - Churners are users who stop using a given service after previously signing up. In the domain of telecommunications and video games, churners represent a loss of revenue as a user leaving indicates that they will no longer pay for the service. In the context of online community platforms (e.g., community message boards, social networking sites, question--answering systems, etc.), the churning of a user can represent different kinds of loss: of social capital, of expertise, or of a vibrant individual who is a mediator for interaction and communication. Detecting which users are likely to churn from online communities, therefore, enables community managers to offer incentives to entice those users back; as retention is less expensive than re-signing users up. In this article, we tackle the task of detecting churners on four online community platforms by mining user development signals. These signals explain how users have evolved along different dimensions (i.e., social and lexical) relative to their prior behaviour and the community in which they have interacted. We present a linear model, based upon elastic-net regularisation, that uses extracted features from the signals to detect churners. Our evaluation of this model against several state of the art baselines, including our own prior work, empirically demonstrates the superior performance that this approach achieves for several experimental settings. This article presents a novel approach to churn prediction that takes a different route from existing approaches that are based on measuring static social network properties of users (e.g., centrality, in-degree, etc.).
U2 - 10.1145/2798730
DO - 10.1145/2798730
M3 - Journal article
VL - 10
JO - ACM Transactions on Knowledge Discovery from Data
JF - ACM Transactions on Knowledge Discovery from Data
SN - 1556-472X
IS - 3
M1 - 21
ER -