Revealing Life Events from Inferred Customer Similarity: A Predictive Modeling Approach

  • Enric Junque de Fortuny
  • Tingting Nian
  • Foster Provost

Is there a way to infer relocation, divorce or other life events from consumer behavior?  This paper develops a predictive model based on the principle that consumer similarity derived from past transaction behavior reveals similarity in life events as well. In this project, we use a unique data set from a Fortune 500 mutual financial institution and combine various heterogeneous features in a predictive model to answer these and related questions.  The main goal of our predictive model is to pro-actively detect life events from consumer demographics and online transaction behavior.  We combine both the state-of-the-art socio-demographic predictive modeling with inferred consumer similarity based on transaction patterns-consumers are linked if they perform similar online transactions.  The results demonstrate that the inferred consumer similarity is a key feature in predicting major life events.