Cost-sensitive churn prediction in fund management services

Abstract

Churn prediction is vital to companies as to identify potential churners and prevent losses in advance. Although it has been addressed as a classification task and a variety of models have been employed in practice, fund management services have presented several special challenges. One is that financial data is extremely imbalanced since only a tiny proportion of customers leave every year. Another is a unique cost-sensitive learning problem, i.e., costs of wrong predictions for churners should be related to their account balances, while costs of wrong predictions for non-churners should be the same. To address these issues, this paper proposes a new churn prediction model based on ensemble learning. In our model, multiple classifiers are built using sampled datasets to tackle the imbalanced data issue while exploiting data fully. Moreover, a novel sampling strategy is proposed to deal with the unique cost-sensitive issue. This model has been deployed in one of the leading fund management institutions in Australia, and its effectiveness has been fully validated in real applications.