Used Car Price Prediction – ML API Deployment

Background
In the second-hand car market, determining a fair market price is crucial for both sellers and buyers. This project aims to leverage machine learning to predict used car prices based on various attributes such as brand, mileage, transmission, fuel type, and more.
The Problems
- Manual price estimation is inconsistent and subjective
- High price variability across similar car listings
- Lack of structured pricing tools for individual sellers
Our Approach
We collected historical car listings using web scraping from a popular Indonesian used car e-commerce website. After performing exploratory data analysis and preprocessing, we trained a regression model to learn from relevant features. The trained model was deployed via a RESTful API using FastAPI, enabling easy access to the prediction service.
Project Showcase

Explore the full ML workflow from data collection to API deployment.
Methodology & Key Features
- Python-based web scraping using BeautifulSoup and Selenium
- EDA and feature engineering with Pandas, Seaborn, and Scikit-learn
- Model selection and evaluation using MAE and RMSE metrics
- Deployment via API
- JSON-based API response for easy frontend integration
- Integrated with a simple UI form to test predictions
Results
The deployed model achieved an MAE of less than IDR 5 million and provided consistent price estimates across various car brands and conditions. The API was used by internal users to evaluate pricing scenarios quickly.
Benefits
- Fast, automated price estimation
- Improved buyer-seller transparency
- Reusable API for future frontend or mobile apps
- Scalable backend ready for marketplace integration