|
|
--- |
|
|
title: FipeFinder AI - Car Price Prediction |
|
|
emoji: 🚗 |
|
|
colorFrom: blue |
|
|
colorTo: green |
|
|
sdk: gradio |
|
|
sdk_version: 5.35.0 |
|
|
python_version: '3.11' |
|
|
app_file: app.py |
|
|
pinned: false |
|
|
--- |
|
|
|
|
|
# 🚗 FipeFinder AI: Used Car Price Predictor |
|
|
|
|
|
This is a Machine Learning application that predicts the selling price of used cars in the Brazilian market. Fill in the vehicle's features to get a price estimate generated by a fine-tuned Transformer model. |
|
|
|
|
|
--- |
|
|
|
|
|
## 🛠️ How It Works |
|
|
|
|
|
This application is powered by a Deep Learning regression model. The main components are: |
|
|
|
|
|
* **Model:** A `BERTimbau` (BERT for Portuguese) model was fine-tuned on a dataset of over 500,000 real car listings from Brazil. |
|
|
* **Feature Engineering:** The model was trained not just on raw data, but on engineered features like vehicle age, usage rate (km/year), and brand/model popularity. |
|
|
* **Training:** The model was trained using PyTorch with GPU acceleration (NVIDIA CUDA) to handle the large dataset efficiently. |
|
|
* **Interface:** The user interface is built with **Gradio**, allowing for easy and interactive predictions. |
|
|
|
|
|
The complete data analysis, feature engineering, and model training process is documented in the project's main notebook. |
|
|
|
|
|
* **[View the full Training Notebook]([#](https://colab.research.google.com/drive/1ZyobI02zJu86NTjUZ043ACGBJn3S3Wuy?usp=sharing))** --- |
|
|
|
|
|
## 🚀 Tech Stack |
|
|
|
|
|
* **Machine Learning:** PyTorch, Hugging Face Transformers, Scikit-learn |
|
|
* **Data Handling:** Pandas, NumPy |
|
|
* **UI & Deployment:** Gradio, Hugging Face Spaces |
