Loading...

Customer Churn Prediction in the Iran Insurance Industry

Etemad Hosseini, Amir Hossein | 2021

285 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 54244 (44)
  4. University: Sharif University of Technology
  5. Department: Management and Economics
  6. Advisor(s): Aslani, Shirin; Arian, Hamid Reza
  7. Abstract:
  8. Insurance companies in Iran operate in a completely competitive and dynamic environment. Because customer acquisition in these companies is significantly more expensive than customer retention, with timely forecasting of churning customers, they can manage their customers more effectively. In this study, in order to predict customer churn in the insurance industry, the data of one of the Iranian insurance companies that has more than two million insurers were used. In order to identify important data and variables, previous studies were reviewed, and on the other hand, the Central Insurance Regulations of the Islamic Republic of Iran, as well as the information of the insurance contracts of this company, were fully and comprehensively reviewed. Using the literature review, 10 machine learning algorithms that were most used in the customer churn prediction literature were used for modeling. After extracting the data from the database, we pre-processed the data using the Python programming language and then trained and tested the models. Finally, it was observed that the Naive Baye model with 95.17% recall has the best performance according to this evaluation metric. Then, in order to extract insights from the models, logistic regression, decision tree and extreme gradient boosting were examined and it was found that the variables including insurance diversity, number of insurance policies of one year ago, number of insurance policies of two years ago, contract, has life insurance and discount percent are the most important variables in predicting customer churn in this insurance company
  9. Keywords:
  10. Customer Churn Prediction ; Imbalanced Data ; Evaluation Method ; Data Mining ; Insurance Industry ; Machine Learning

 Digital Object List

 Bookmark

No TOC