car-price-prediction-model

Car Price Prediction Model

Project Overview

This project aims to predict a car’s market price (MSRP) based on technical and categorical vehicle features.

Goal

Predict a car’s price using machine learning models and engineered features.

Business Value

Help buyers and sellers estimate fair car prices and understand key price drivers.

---

Dataset

• ~11,000 rows
• 16 original features
• Includes both numerical and categorical variables
• Source: Kaggle (Car Features and MSRP dataset)

---

Data Processing

• Removed duplicates
• Handled missing values:
  • Numerical → median
  • Categorical → mode / "Unknown"
• Log transformation applied to target variable (MSRP)
• Feature engineering:
  • car_age
  • hp_per_cylinder
  • mpg_combined
• One-hot encoding for categorical features

---

Exploratory Data Analysis

• Strong right skew in price distribution → log transformation applied
• Strong correlation found between:
  • Engine HP and price
  • Year and price
• Visual analysis showed:
  • Higher horsepower → higher price
  • Newer cars → higher price

---

Models Used

• Linear Regression
• Random Forest Regressor
• Gradient Boosting Regressor

---

Results

Best Model: Random Forest

---

Feature Importance Insights

Top features:

• Car Age
• Year
• Engine HP

Conclusion: Vehicle depreciation and performance are the most important factors in determining price.

---

Model File

The trained model is saved as: car_price_model.pkl

---

Future Improvements

• Hyperparameter tuning
• Advanced models (XGBoost, LightGBM)
• Deployment as API / web app

---

Author

Tal Cabalo