PEER-TO-PEER LENDING LOAN GRADE AND INTEREST RATE PREDICTION USING APPLICANT FINANCIAL SIGNAL

Abstract

Medical insurance cost prediction plays a significant role in helping insurance companies
estimate healthcare expenses based on individual health and demographic factors. This
research proposes a machine learning-based Medical Insurance Cost Prediction System that
predicts insurance premiums using attributes such as age, gender, body mass index (BMI),
number of children, smoking status, and region. The dataset is first preprocessed by
handling categorical variables using label encoding and applying feature engineering
techniques such as BMI–Smoker and Age–BMI interactions to improve model
performance.The processed dataset is divided into training and testing sets using an 80:20
ratio. A Random Forest Regression model is then trained to learn the relationship between
input features and insurance charges. The model achieves an accuracy of approximately 91
– 92% based on R² score, indicating good predictive capability. Experimental results show
that the system can accurately estimate medical insurance costs. For example, for inputs
Age = 40, BMI = 20.9, Children = 2, Smoker = No, the predicted insurance cost is ₹7,788.05.
This system can assist insurance providers in automated premium estimation.
This project uses machine learning techniques to predict insurance charges based on
factors such as age, BMI, smoking status, gender, and region