Survival Analysis


Survival analysis is often associated with the medical field, particularly in treatment trials where researchers are predicting how long a patient will survive or continue in remission, considering a certain treatment plan. More generally, survival analysis models predict how long before a particular event occurs. In many domains, where simple classification has been used to predict the occurrence of an event, survival analysis can be used to add the valuable dimension of time until that occurrence.

Credit Risk Analysis

Credit scoring generally falls into two categories: application and behavior scoring. The first uses the customer data from the time of application as well as information about the loan to predict whether the customer will default on the loan. After a loan is granted, the creditor begins gathering information about the customer’s behavior, such as payments or spending. Survival analysis models incorporate this behavior data into a model to predict not only if a particular account will default, but how long before that happens. This time component allows financial institutions to prepare for upcoming conditions, predicting potential losses more accurately and allowing better use of capital. In some cases, a default after a certain period is still a profitable account, so the traditional good/bad classification is insufficient in terms of financial planning.

Churn Prediction

Churn is defined as customers who leave a company. Some churn is unavoidable, ie a person moves out of a service area, and not so interesting to the company. Predicting other types of churn can be quite profitable to a company as the costs of maintaining a customer are often much less than recruiting new customers. Traditional churn prediction is based on binary classification, predicted to churn or predicted not to churn. Survival analysis, as it does in other domains, predicts not only if a customer will churn but how long until they are expected to churn. This allows a company to intervene with some incentives for the customer to stay with the company. Accurate prediction of upcoming churners results in highly-targeted campaigns, limiting the resources spent on customers who likely would have stayed anyway.

Notable Publications

  • Louis, P.Van Laere, E.Baesens, B. (2011). Motivating and predicting bank rating transitions using optimal survival analysis modelsProceedings of the 24th Australasian Finance & Banking Conference: Vol. acceptedAustralasian Finance & Banking ConferenceSydney (Australia)14-16 December 2011.
  • Louis, P.Van Laere, E.Baesens, B. (2011). Predicting bank rating transitions using optimal competing risks survival analysis modelsProceedings of the credit scoring and credit control XII conference: Vol. acceptedCredit Scoring and Credit Control conferenceEdinburgh (UK)24-26 August 2011.
  • Baesens, B.Van Gestel, T.Stepanova, M.Vanthienen, J. (2003). Neural network survival analysis for personal loan dataProceedings of the Eighth Conference on Credit Scoring and Credit Control (CSCCVII’2003)Conference on Credit Scoring and Credit Control (CSCCVII’2003)Edinburgh (Scottland).
  • Tsujitani, M., Baesens, B. (2011). Survival analysis for personal loan data using generalized additive models. Behaviormetrika39 (1), 9-23.
  • Baesens, B.Van Gestel, T.Stepanova, M.Vanthienen, J.Van den Poel, D. (2005). Neural network survival analysis for personal loan dataJournal of the Operational Research Society56 (9 (Sept.))1089-1098.