To predict the ultimate individual claim amounts on historical FNOL Motor Data
Problem Statement: Develop a predictive model leveraging FNOL (First Notification Of Loss) characteristics from a historical dataset of head-on collision insurance claims. The model aims to estimate the future settlement costs of individual claims, enabling more accurate and proactive pricing decisions for insurance company.
Solution: To build an ingestion pipeline of data pre-processing steps to translate an incoming raw FNOL dataset for modeling and training needs which includes data cleaning, data transformation (encoding categorical), feature engineering and feature selection. Train a global XGBoost Model and set up model deployment to make predictions on future FNOL data.
Key Takeaways: By providing early estimation of ultimate incurred loss and identifying top features, it improves the ability of transparent data driven decision-making, better risk management and allocation of resources. Company could also realize Cost Savings and Competitive Advantage by offering more precise fairy price insurance products and better claims handling which will ultimately improving customer satisfaction.