Complete Schedule

Presentation Type

Open Access Oral Presentation

Abstract

  • Insurance companies rely heavily on accurate risk prediction to identify potential customers, make marketing easier, and reducing the chances of loosing money.
  • Traditionally, actuarial models such as logistic and Probit regression have been widely used because they are interpretable and statistically solid.
  • However because of the rapid growth of data, insurance companies have started to shift from these traditional methods to machine learning techniques that can capture more complex and nonlinear relationships which traditional models may miss
  • Therefore, this study investigates whether modern machine learning techniques—specifically Random Forest and XGBoost—can outperform traditional actuarial models in predicting customer interest in vehicle insurance.
  • In addition to comparing predictive performance, the study also analyzes which variables are most influential across models.
  • This helps insurance companies to identify what influences customers’ decisions and what factors encourage them to purchase insurance.”

Start Date

14-4-2026 2:30 PM

End Date

14-4-2026 2:45 PM

Faculty Advisor

Douglas Mupasiri

Department

Department of Mathematics

Student Type

Undergraduate Student

File Format

application/pdf

File Size

1.01 MB

Download

Comments?

Share

COinS
 
Apr 14th, 2:30 PM Apr 14th, 2:45 PM

AI in Insurance Industry: Using Machine Learning to Improve Insurance Risk Predictions

  • Insurance companies rely heavily on accurate risk prediction to identify potential customers, make marketing easier, and reducing the chances of loosing money.
  • Traditionally, actuarial models such as logistic and Probit regression have been widely used because they are interpretable and statistically solid.
  • However because of the rapid growth of data, insurance companies have started to shift from these traditional methods to machine learning techniques that can capture more complex and nonlinear relationships which traditional models may miss
  • Therefore, this study investigates whether modern machine learning techniques—specifically Random Forest and XGBoost—can outperform traditional actuarial models in predicting customer interest in vehicle insurance.
  • In addition to comparing predictive performance, the study also analyzes which variables are most influential across models.
  • This helps insurance companies to identify what influences customers’ decisions and what factors encourage them to purchase insurance.”