Medical_Premium_Price_Prediction

Medical Premium Price Prediction – End-to-End ML Project with Deployment

Deployed on Streamlit: View Live App

---

Table of Contents

• Project Overview
• Dataset
• Data Exploration & Cleaning
• Data Preprocessing
• Model Selection
• Hyperparameter Tuning
• Final Model Performance
• Predictions
• Key Takeaways
• Technologies & Libraries Used

---

Project Overview

Predict the insurance premium cost for individuals based on their demographic and medical history.
The goal is to estimate accurate premium prices, helping both insurance providers and customers make informed decisions.

---

Dataset

• File: Medicalpremium.csv
• Rows: ~1,000
• Columns: 11 features + target (Insurance Premium)

Features

• Age
• Diabetes (0/1)
• Blood Pressure Problems (0/1)
• Transplants (0/1)
• Chronic Diseases (0/1)
• Height (cm)
• Weight (kg)
• Known Allergies (0/1)
• Cancer History in Family (0/1)
• Number of Major Surgeries
• Insurance Premium (Target)

---

Data Exploration & Cleaning

• Checked for null values → none found
• Verified data ranges:
  • Age (18–100)
  • Height (100–220 cm)
  • Weight (30–200 kg)
• Converted categorical features (Diabetes, Blood Pressure, etc.) to numeric 0/1
• No duplicate entries found

---

Data Preprocessing

• Feature Scaling: Applied StandardScaler on numerical inputs (Age, Height, Weight, etc.)
• Saved scaler as scaler.pkl for deployment consistency

---

Model Selection

Models evaluated:

• Linear Regression
• Decision Tree Regressor
• Random Forest Regressor

Metrics: MAE, MSE, RMSE, R² Score

Random Forest performed best among all models

---

Hyperparameter Tuning

Used GridSearchCV with 5-fold cross-validation

Best Parameters:

{
 'criterion': 'squared_error',
 'max_depth': 8,
 'max_features': 'sqrt',
 'min_samples_leaf': 1,
 'min_samples_split': 5,
 'n_estimators': 100
}

Final Model Performance (Test Split)

Metric	Random Forest
MAE	~665
MSE	~740,324
RMSE	~860
R²	~0.0035

Predictions

• User enters health details (age, conditions, surgeries, etc.) in the Streamlit app
• Input is scaled using saved scaler.pkl
• Model (best_model.pkl) predicts the insurance premium amount

Key Takeaways

• Random Forest outperformed other regression models
• Scaling features ensures consistent predictions
• Hyperparameter tuning improves model stability
• Deployment on Streamlit allows interactive real-time estimation

Technologies & Libraries Used

• Programming Language: Python 3.10

Libraries:

• pandas – data manipulation
• numpy – numerical operations
• scikit-learn – preprocessing, model selection, evaluation
• matplotlib, seaborn – visualization
• streamlit – web app deployment