Car Price Prediction with Machine Learning

Project Overview

This project uses machine learning algorithms to predict car prices based on various features such as engine specifications, fuel type, body style, and more. The model is trained on a dataset containing detailed information about different car models and their corresponding prices.

Dataset

Source: Kaggle - Car Price Prediction Dataset

The dataset contains information about cars with the following key features:

• Engine specifications: Engine size, horsepower, fuel system
• Physical attributes: Length, width, height, curb weight
• Performance metrics: City MPG, highway MPG, compression ratio
• Categories: Car company, fuel type, aspiration, body style, drive wheels
• Target variable: Price (what we're predicting)

Dataset Files

• CarPrice_Assignment.csv - Main dataset with car features and prices
• Data Dictionary - carprices.xlsx - Detailed description of all features

Technologies Used

• Python 3.x
• pandas - Data manipulation and analysis
• numpy - Numerical computing
• scikit-learn - Machine learning algorithms and tools
• matplotlib - Data visualization
• Custom transformers - Log transformation for specific features

Project Structure

PredictCarPrice/
│
├── dataset/
│   ├── CarPrice_Assignment.csv       # Main dataset
│   └── Data Dictionary - carprices.xlsx  # Feature descriptions
│
├── main.py                          # Main script with model training
├── loadData.py                      # Data loading and splitting utilities
├── dataCleanup.py                   # Data preprocessing and feature engineering
├── custom_transformers.py           # Custom sklearn transformers
├── .gitignore                       # Git ignore file
└── README.md                        # This file

Features & Data Processing

Feature Engineering

• Log transformation applied to horsepower and enginesize for better distribution
• Robust scaling for numerical features to handle outliers
• One-hot encoding for categorical variables
• Custom pipeline for seamless data preprocessing

Data Preprocessing Pipeline

1. LogTransformer: Applies log transformation to specified numerical features
2. RobustScaler: Scales numerical features while being robust to outliers
3. OneHotEncoder: Converts categorical variables to numerical format

Machine Learning Models

The project evaluates multiple regression algorithms:

1. Linear Regression - Baseline model
2. Decision Tree Regressor - Non-linear relationships
3. Random Forest Regressor - Ensemble method (final choice)

Model Selection Process

• Cross-validation with 10 folds for robust evaluation
• Grid Search for hyperparameter tuning
• RMSE (Root Mean Square Error) as the primary evaluation metric

Hyperparameter Tuning

The Random Forest model is optimized using GridSearchCV with the following parameters:

• n_estimators: [3, 10, 30]
• max_features: [2, 4, 6, 8]
• bootstrap: [True, False]

How to Run

1. Clone the repository:

   git clone https://github.com/ArnavGRao/PredictCarPriceWithRegression.git
   cd PredictCarPriceWithRegression

2. Install required packages:

   pip install -r requirements.txt

   Or install manually:

   pip install pandas numpy scikit-learn matplotlib scipy seaborn

3. Run the main script:

   python main.py

Key Functions

main.py

• Main execution script
• Model training, evaluation, and comparison
• Hyperparameter tuning with GridSearchCV
• Final model evaluation and feature importance analysis

loadData.py

• Data loading utilities
• Train-test split functionality

dataCleanup.py

• Data preprocessing functions
• Feature engineering pipeline

custom_transformers.py

• LogTransformer: Custom sklearn transformer for log transformation

Model Performance

The project includes comprehensive model evaluation:

• Cross-validation scores for model reliability
• Grid search results for optimal hyperparameters
• Feature importance rankings to understand which car attributes drive price predictions
• Test set evaluation for final model performance assessment

Key Learnings

• End-to-end ML pipeline from data loading to model evaluation
• Proper data preprocessing techniques for mixed data types
• Model comparison and selection methodology
• Hyperparameter optimization using grid search
• Custom transformer creation for specialized preprocessing needs

Author

ArnavGRao

• GitHub: @ArnavGRao
• Email: arnavgrao@gmail.com

License

This project is open source and available under the MIT License.