I'm Marcos Mansur, a brazilian production engineering student very passionate about programming, data science and machine learning!
Strongly problem solving oriented and motivated by challenges and puzzles. Very curious, always learning!
You can contact me through Linkedin in the following link: https://www.linkedin.com/in/mmansur/
Send me a message, let's talk and connect!
If you can read portuguese, I write about machine learning and Data Science at: https://medium.com/@mmansur
-
Titanic survival prediction model - Machine Learning model to predict survivability of Titanic passangers. CI basic structure on yaml file to print performance metrics as comments in pull requests. This model scores TOP 3% at "Titanic - Machine Learning from Disaster" Kaggle competition.
-
Time series Regression Model for VLabs Challange - ElasticNet Regressor to predict how much each client is going to spend in the next 90 days based on data from 14 months of purchases. 5 diferent models to predict customer lifetime value by channel of sale. Scored second place in the leaderboard out of 17 teams. (Notebooks in portuguese-BR)
-
Tabular Data Sep21 Analysis- Kaggle competition - LightGBM regressor tuned with bayesian optimization (optuna) for the tabular data competition. The dataset has 118 unnamed numeric features and the missing values are significantly relevant for prediction.
-
IDHM prediction of brazilian cities - Voting regressor of 4 tree-based algorithms to predict de IDHM (Municipal Human Development Index) of brazilian cities
-
Deployed Web App to compare routes on a map - A Streamlit based deployed web app that plot multiple routes on a map for comparison and geospatial analysis. You can play with the app at: https://share.streamlit.io/marcos-mansur/plot_route_app/main/MyApp.py
-
Web Scrapping learning projects - a robot to download book covers from https://inventwithpython.com/ and another that plays simple instructions and record the board of each play at 2048 with selenium and bs4
-
Study notes - Notes from the early studies of SQL, Python and machine learning (basic commands, pandas, numpy) and web scrapping using selenium and bs4
-
Code Wars katas - solutions to a few katas from Code Wars ranging from 8 kyu to 5 kyu