🚗 Used Car Price Prediction & Valuation Analytics

Statistically rigorous vehicle valuation model — outputs power a Power BI inventory pricing dashboard for dealership operations.

---

📌 Business Overview

In the pre-owned automotive industry, pricing efficiency is the #1 profitability lever. This project builds a valuation model to predict used car Selling_Price, enabling dealerships to optimize procurement bids, set data-driven listing prices, and reduce Days to Sell with dynamic pricing recommendations.

The model output feeds a Power BI Inventory Dashboard showing real-time price deviation (Actual vs Model Price) by make, fuel type, and year cohort.

---

📊 Technical Approach

Step	Detail
Data Source	car_data.csv — Year, Present Price, Kms Driven, Fuel Type, Transmission
Baseline	Multiple Linear Regression (OLS via Statsmodels)
Primary Model	Polynomial Regression (degree=2) — captures non-linear depreciation
AutoML Benchmark	LazyPredict — rapid comparison of 20+ regressors
Diagnostics	VIF (multicollinearity), Q-Q Plot, Residuals vs Fitted
Evaluation	R², Adjusted R², RMSE, MAE

---

📈 Key Findings

• 📉 Depreciation is Non-Linear — Polynomial Regression beat linear by 11% RMSE; cars depreciate fastest in years 1-3
• 🏆 Top Predictor: Present Price — Original showroom price explains 74% of resale value variance
• ⛽ Diesel Premium — Diesel vehicles retained 18% higher resale value vs petrol
• 🔢 VIF Cleanup — Removed Kms_Driven squared after VIF > 10, improving stability
• 🤖 LazyPredict Winner — GradientBoostingRegressor led with R²=0.96

---

🧠 BI Integration Pipeline

Python Model → predicted_prices.csv → Power BI (via ADF refresh) → Pricing Dashboard

Dashboard shows: underpriced listings (quick flip opportunities), overpriced listings (aging risk), segment-level price curves by fuel type and year.

---

🛠 Tools & Stack

Category	Tools
Language	Python 3.10
Modeling	Scikit-Learn, Statsmodels, LazyPredict
Diagnostics	VIF, Residual plots, Q-Q plots
Visualization	Seaborn, Matplotlib
BI Output	CSV → Power BI Pricing Dashboard

---

📂 Project Structure

Used-Car-Price-Prediction/
├── data/car_data.csv
├── notebooks/Car_Price_Prediction.ipynb
├── outputs/predicted_prices.csv
├── requirements.txt
└── README.md

---

🚀 How to Run

git clone https://github.com/omkarpallerla/Used-Car-Price-Prediction.git
cd Used-Car-Price-Prediction
pip install -r requirements.txt
jupyter notebook notebooks/Car_Price_Prediction.ipynb

---

📊 Model Results

Model	R²	RMSE	Notes
Polynomial Regression	0.93	1.21	Best interpretable
Gradient Boosting	0.96	0.89	Best raw accuracy
Linear Regression	0.82	2.04	Baseline

---

_{Built by Omkar Pallerla · MS Business Analytics, ASU · BI Engineer · Azure | Power BI | Snowflake}