Open Universiteit

Please use this identifier to cite or link to this item: http://hdl.handle.net/1820/7589
Title: Churn prediction models tested and evaluated in the Dutch indemnity industry
Authors: Siemes, T
Keywords: Data mining
AUC
ROC
confusion matrix
support vector machine
neural networks
decision tree
logistic regression
comparison
churn prediction
customer churn
indemnity insurance industry
Issue Date: 1-Nov-2016
Publisher: Open Universiteit Nederland
Abstract: Due to global developments customer churn is getting a growing concern to the insurance industry. Technological improvements like the internet makes it much easier for customer to compare their policies, obtain new offers or even churn from one provider to another. The insurance industry therefore has become a heavily competitive market in which insurance companies have to compete to protect and expand their customer base in order to maintain or expand their market position. Thus, retaining customers is becoming more and more important and therefore finding customers who are most likely to leave is a central aspect. Many different techniques are available to identify customers who are most likely to leave, however which technique can be used best is often not clear. Research clarifies that the characteristics of the industry and/or dataset which is used are mostly assessing related to performance. In advance it is impossible to determine the best suited technique to use if previous research in which performance was tested has not been published. This study presents a data mining methodology in which the four most used prediction techniques in literature are tested and evaluated using a real life voluminous insurance company dataset to determine which technique performs best. Using the same dataset makes results comparable and clears out which technique performs best based on the insurance data domain characteristics.
URI: http://hdl.handle.net/1820/7589
Appears in Collections:MSc Management Science

Files in This Item:
File Description SizeFormat 
Siemes T scriptie dspace.pdf2.13 MBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.