vehicle-price-predictor

Vehicle Price Prediction and Market Analysis

This project analyzes the Australian Vehicle Prices dataset (from Kaggle) and builds multiple models to:

Understand the key factors influencing car prices

Predict vehicle prices using regression models

Segment vehicles into meaningful clusters

Classify vehicles into Standard vs Premium categories

The project includes extensive EDA, feature engineering, regression modeling, clustering, and classification.

1. Dataset Overview

The dataset includes 16,774 vehicle listings from the Australian market, containing attributes such as:

Brand, Model

Year

Kilometres driven

Engine size

Fuel consumption

Transmission

Drive type

Doors, Seats

Body type

Color

Price (target variable)

Research Question: What are the key factors that influence vehicle prices in the Australian market, and how accurately can we predict a car’s price?

1. Data Cleaning & Feature Engineering Steps Performed:

Removed listings with missing Price

Filled missing categorical values with "Unknown"

Converted text fields such as "4 doors" → numeric values

Converted "5 seats" → numeric values

Created new features:

Car_Age = 2024 – Year

Kilometres_Clean (outlier-adjusted)

Efficiency = Engine_Liters / FuelConsumption

Is_Luxury based on known luxury brands (e.g., BMW, Mercedes, Porsche)

1. Exploratory Data Analysis (EDA)

Below are the included graphs as Markdown placeholders.

3.1 Price Distribution

3.2 Kilometres Distribution

3.3 Engine Size Distribution

3.4 Fuel Consumption Distribution

3.5 Price vs Kilometres

3.6 Price vs Year

3.7 Price Distribution by Brand

3.8 Price by Drive Type

3.9 Price by Doors

3.10 Price by Seats

3.11 Price by Body Type

3.12 Price by Transmission Type

3.13 Correlation Heatmap

Key Insights from EDA:

Newer vehicles and low-kilometre vehicles are significantly more expensive.

Engine size and number of cylinders strongly correlate with price.

Luxury brands (Porsche, Lamborghini, Ferrari, McLaren) show extreme right-skewed distributions.

SUVs, Coupes, and Premium categories have much higher median prices.

Fuel efficiency has a mild negative correlation with price (larger engines consume more fuel and tend to be pricier).

1. Regression Modeling

Three models were trained:

Models:

Linear Regression

Random Forest Regressor

XGBoost Regressor

4.1 Actual vs Predicted Prices

4.2 Residuals Plot

4.3 Feature Importance (Linear Regression)

Key Regression Findings:

Linear Regression underfits high-priced vehicles due to non-linearity.

Random Forest and XGBoost perform significantly better with nonlinear interactions.

Cylinders, Engine Size, Year, and Kilometres are the top predictors of price.

1. K-Means Clustering (Unsupervised Learning) K=3 was selected based on interpretability. 5.1 PCA Visualization of Clusters

Cluster Interpretation: Cluster Characteristics 0 Older, lower engine size, lower price 1 Mid-range vehicles, balanced features 2 Newer, powerful engines, premium pricing 6. Advanced Modeling: Random Forest & XGBoost 6.1 Performance Comparison

R² Scores:

Random Forest: ~0.89

XGBoost: ~0.83

Linear Regression: ~0.67

6.2 Feature Importance (Random Forest)

Important Features:

Car Age

Kilometres

Cylinders

Engine Liters

Efficiency

Luxury Brand Indicator

Drive Type (AWD / FWD)

1. Classification Model: Standard vs Premium Vehicles

Created binary label:

0 = Standard vehicle

1 = Premium vehicle

7.1 Class Distribution

7.2 Confusion Matrices

Decision Tree:

KNN:

Random Forest Classifier:

Classification Findings:

Random Forest achieved the highest accuracy and best balance across classes.

Premium vehicles are harder to classify due to smaller sample size (class imbalance).

Important features for classification included: Cylinders, Engine Size, Luxury Brand Indicator, Drive Type, Cluster ID.

1. Conclusions & Insights Market Insights:

Price is most strongly driven by engine size, cylinders, year, kilometres, and luxury brand.

SUVs and Coupes show consistently higher price ranges.

Luxury brands dominate the top of the market with extreme price outliers.

Modeling Insights:

Simple linear models cannot capture the nonlinear structure of car prices.

Random Forest provides the best regression performance (R² ≈ 0.89).

Clustering reveals meaningful market segments.

Classification models perform well but are limited by class imbalance.

video: https://www.loom.com/share/f157870be4a6463a8bb10624ff12f218