×

Geodata Web App (Machine Learning)

A solution to help businesses find the best locations for their international expansion, based on population density data and machine learning (KMeans Clustering).

This app was built as a final project for the Le Wagon Data Science bootcamp. It was was built in a team of four, which I had the honor to lead. This is the tech stack chosen for this project:

Technologies

• Python
• FastAPI
• PostgreSQL
• Google Cloud Platform
• Docker
• Streamlit
• Heroku

My contributions

• Project management
• Data sourcing
• Model development with Scikit-Learn
• Backend development
• Deployment on Google Cloud Platform

×

E-Commerce Case Study

Case Study from the Olist Kaggle Challenge

Olist is a Brazilian e-commerce B2B platform.

The problem: How can we increase profit margin, knowing that bad reviews cause reputation damage?

The dataset: 120 MB of data containing 100,000 real orders

The task: Analyze what causes negative reviews and suggest appropriate action, assuming that negative reviews have a measurable "cost" of 40-100 BRL (7-18 USD) per review.

Employed methods:

• Statistical modeling using Logistic Regression
• Hypothesis testing
• Data analysis using Python Pandas, SciPy, and Statsmodels
• Data visualization using Python Matplotlib, Seaborn, and Plotly

Results:

The results of the analysis were summarized in the following presentation:

Results summary