Optimizing Motor Insurance Premiums in the UAE Using Predictive Analytics
Saved in:
| Title: | Optimizing Motor Insurance Premiums in the UAE Using Predictive Analytics |
|---|---|
| Authors: | Abdelsalam, Samar |
| Committee Members: | Ioannis Karamitsos |
| Summary: | The car insurance industry in the UAE and GCC faces growing challenges including premium inflation, fraudulent claims, inconsistent pricing models, and lagging innovation in underwriting. This study investigates how predictive analytics and machine learning can enhance motor insurance premium optimization, with a focus on incorporating socio-demographic, behavioral, environmental, and vehicle-specific risk factors. Using a publicly available dataset of over 125,000 insurance policy records, this research applies a CRISP-DM framework to develop and evaluate supervised learning models including Linear Regression, Decision Trees, and Random Forests. Results highlight Linear Regression as the most effective model (RMSE = 193.62, R² = 0.9016), enabling accurate premium estimation and improved pricing fairness. Complementary qualitative analysis contextualizes findings with industry-specific challenges such as regulatory constraints and customer demand for personalization. The project delivers a predictive modeling tool, interactive dashboards for risk profiling, and strategic recommendations for insurers aiming to implement data-driven premium structures. Ultimately, this research supports a transition toward more transparent, efficient, and equitable insurance practices in the UAE and GCC. |
| URL: | https://repository.rit.edu/theses/12523 |
| Database: | OpenDissertations |
| Abstract: | The car insurance industry in the UAE and GCC faces growing challenges including premium inflation, fraudulent claims, inconsistent pricing models, and lagging innovation in underwriting. This study investigates how predictive analytics and machine learning can enhance motor insurance premium optimization, with a focus on incorporating socio-demographic, behavioral, environmental, and vehicle-specific risk factors. Using a publicly available dataset of over 125,000 insurance policy records, this research applies a CRISP-DM framework to develop and evaluate supervised learning models including Linear Regression, Decision Trees, and Random Forests. Results highlight Linear Regression as the most effective model (RMSE = 193.62, R² = 0.9016), enabling accurate premium estimation and improved pricing fairness. Complementary qualitative analysis contextualizes findings with industry-specific challenges such as regulatory constraints and customer demand for personalization. The project delivers a predictive modeling tool, interactive dashboards for risk profiling, and strategic recommendations for insurers aiming to implement data-driven premium structures. Ultimately, this research supports a transition toward more transparent, efficient, and equitable insurance practices in the UAE and GCC. |
|---|
