Mangio

Inspiration

Mangio was developed during a datathon organized by UPC – Universitat Politècnica de Catalunya, and our team chose to take on the Mango challenge, aimed at finding innovative solutions for production optimization. As beginner data scientists, this was a unique opportunity to apply our skills to a real-world business problem.

What It Does

Mangio predicts production outputs for Mango, enabling better planning and resource allocation. By analyzing historical data, our model identifies patterns and generates reliable forecasts for business decision-making.

How We Built It

We used Python for data manipulation and model building, leveraging Pandas and NumPy for preprocessing and feature engineering. The predictive model was implemented with LightGBM and optimized using Optuna for hyperparameter tuning. Jupyter Notebook allowed us to collaborate effectively and visualize results clearly.

Challenges We Ran Into

Being a team of novices, almost every step—from data cleaning to model tuning—was a learning experience. Coordinating our workflow, ensuring data quality, and maintaining reproducible experiments were key challenges.

Accomplishments We're Proud Of

We built a complete end-to-end data pipeline for the Mango challenge, including filtering, preprocessing, model training, and evaluation. Beyond technical results, the teamwork, dedication, and rapid learning demonstrated by our team were major achievements.

What We Learned

We gained hands-on experience in LightGBM, hyperparameter tuning, and structured data preprocessing. Additionally, we strengthened our collaboration, project management, and version control skills—essential for real-world data science projects.

What's Next for Mangio

We plan to refine the model further, explore additional data sources, and improve prediction accuracy. This datathon has inspired us to continue advancing our skills and tackle more complex business challenges.