Car Dheko - Used Car Price Prediction Application

This project is a Streamlit-based web application that predicts the price of used cars based on a comprehensive analysis of structured and unstructured car listing data scraped from CarDekho. By implementing a robust preprocessing pipeline and a fine-tuned Random Forest model with GridSearchCV, the application offers users accurate(R² score 0.9028) and data-driven car price estimates based on attributes such as city,fuel type, body type, transmission, manufacturer, and more.

Watch Demo Video

Problem Statement

The "Car Price Prediction Application" addresses the challenge of providing reliable pricing estimates for used cars. As data can be unstructured and inconsistent, the project focuses on cleaning, encoding, and scaling to ensure data integrity, followed by hyperparameter tuning to maximize model accuracy. The final model enables precise pricing predictions to support both car buyers and sellers in the used car market.

Key Highlights

• Data Cleaning and Feature Engineering: Handles unstructured data, extracts meaningful features, and ensures consistency across the dataset.
• Optimal Model Selection: Tested multiple models, with Random Forest achieving the best performance:
  • R² Score: 0.895, explaining over 90% of variance in car prices.
  • Mean Absolute Error (MAE): 1.23
  • Mean Squared Error (MSE): 13.12
• Hyperparameter Optimization: Leveraged Grid Search to improve the model's R² score to 0.9028, indicating the model explains approximately 90.28% of price variance.

Business Use Cases

• Used Car Dealerships: Provide data-driven pricing for better customer interactions.
• Private Sellers and Buyers: Help individual users make informed pricing decisions.
• Market Analysis: Gain insights into factors affecting car prices.

Features

1. Data Preprocessing

• Unstructured Data Processing: Transfored data scraped from CarDekho into a structured format.
• Feature Engineering: Extracts critical attributes from complex, nested data fields.
• Scaling and Encoding: Utilizes MinMax Scaler for numerical attributes and one-hot encoding for categorical attributes.
• The preprocessing pipeline is implemented in the Data_preprocessing.ipynb notebook.

2. Model Development

• Model Evaluation: Compared various models, with Random Forest showing the best performance metrics.
• Grid Search Optimization: Enhanced the Random Forest model’s parameters, yielding:
  • R² Score: 0.9028
  • MAE: Significantly reduced from initial testing.
• The model development and training process is detailed in the Car_price_prediction.ipynb notebook.

3. User Interface with Streamlit

• Interactive Web Application: Built with Streamlit for easy user interaction.
• User Inputs: Accepts car features such as fuel type, body type, transmission, manufacturer, and other specifications.
• Prediction Display: Outputs the predicted car price based on the input parameters.
• 
• 
• 

Application Workflow

1. Data Loading and Cleaning: Reads raw data, cleans, and encodes it for analysis.
2. Model Training and Tuning: The Random Forest model is trained, tuned using Grid Search, and evaluated for optimal performance.
3. Price Prediction: The Streamlit app provides an interface for users to enter car features, and the model returns an estimated price.

Technologies Used

• Python: Core language for data processing and modeling.
• Streamlit: Web framework for building the interactive user interface.
• Pandas and NumPy: For data manipulation and preprocessing.
• scikit-learn: For model building, tuning, and evaluation.
• Pickle: For saving the trained model for deployment.

---

Project Files

1. Data_preprocessing.ipynb:

   • Preprocesses the raw car dataset, including feature extraction, cleaning, and encoding.
   • Outputs a structured dataset ready for model training.

2. Car_price_prediction.ipynb:

   • Contains code for training and optimizing the machine learning model.
   • Saves the trained model and scaler as pickle files (car_price_model.pkl and scaler.pkl).

3. app.py:

   • The Streamlit application for user interaction.
   • Loads the trained model and scaler, takes user inputs, and outputs the predicted car price.

---

Model Performance

Model	MAE	MSE	R² Score
Random Forest	1.23	13.12	0.905
Optimized Random Forest	-	-	0.9028

The optimized Random Forest model achieved a high R² score of 0.9028, meaning it explains around 90.28% of the variance in car prices, with reduced MAE and MSE from the initial model version.

---

How to Run

1. Clone the repository:

   git clone <repository_url>
   cd car-price-prediction
   

2. Ensure you have the following libraries installed: streamlit pandas scikit-learn

3. Run the Streamlit Application:

• To run the application, use the following command:

  streamlit run app.py

Contributing:

• Contributions are welcome! Feel free to submit pull requests or open issues for any bugs, feature requests, or improvements.
• Let me know if you need any further changes or assistance!