🚗 Car Price Prediction using Machine Learning

📌 Overview

This repository contains the implementation of a machine learning pipeline for analyzing and predicting second-hand car prices based on various features such as mileage, engine capacity, brand, and fuel type. The project includes regression, classification, and clustering tasks, as well as deep learning techniques for natural language processing (NLP) and computer vision.

📊 Dataset

• Name: Cars Specifications Dataset
• Source: Web scraping
• Size: 146 features including technical, environmental, and performance specifications.

🔍 Tasks Implemented

✔️ Regression: Predicting car prices based on vehicle attributes.
✔️ Classification: Categorizing cars into environmental labels (ECO, C, D).
✔️ Clustering: Grouping cars based on technical features.
✔️ Algorithm Comparison: Evaluating traditional ML vs deep learning models.
✔️ NLP Task: Predicting car price based on textual descriptions.
✔️ Computer Vision: Classifying car brands using images.

🏆 Best Performing Models

Task	Best Model
Regression	Random Forest (lowest MAE & MSE)
Classification	Random Forest (highest accuracy & F1-score)
Clustering	Agglomerative Clustering (highest silhouette score)
NLP	LSTM-based model using BETO embeddings
Computer Vision	Convolutional Neural Network (CNN)

⚙️ Data Processing

• Feature Engineering: Extracted structured features using regex.
• Handling Missing Data: Used regression-based imputation and categorical mode assignment.
• Dimensionality Reduction: PCA and Truncated SVD applied.
• Data Standardization: Normalization and one-hot encoding.

📂 Repository Structure

📂 data/                     # Processed datasets
📂 preprocessing/            # Data cleaning & feature extraction
📂 models/                   # Trained ML & DL models
📂 scripts/                  # Training and evaluation scripts
📂 notebooks/                # Jupyter Notebooks with analysis
📂 utils/                    # Helper functions for preprocessing & visualization

📈 Results & Insights

• Price trends: Car price is strongly influenced by engine power, mileage, and brand.
• Brand value: Some brands retain value better in the second-hand market.
• Model performance: Random Forest consistently outperforms simpler linear models.
• Feature importance: Power, mileage, and environmental labels are key indicators of price.

🔗 References

• 📑 Project Report
• 📊 Data Preprocessing