Project Portfolio

Titanic - Kaggle Competition
The Kaggle 'Titanic - Machine Learning from Disaster' competition is a must-do when it comes to start with ML competitions and familiarize oneself with the Kaggle platform and community.
The competition is simple: use machine learning to create a model that predicts which passengers survived the Titanic shipwreck.
Rank: Top 3% (921st of 37074 as of May 2021)
Link: Titanic Competition

Sydney Rain Forecast
Predict whether or not it will rain tomorrow based on 140.000+ daily weather obervation examples captured all around Australia by the Australian Feredal Bureau of Meteorology (BOM) from 2008 to 2017.
2020 Accuracy: 79% (289 predictions OK / 366)
Link: Rain Tomorrow

Data Science Capstone Project
This project was part of the Johns Hopkins University Data Science Specialization on Coursera. Based on a text dataset containing millions of lines, and while user is typing a text input, the model predicts what the next word is likely to be.
The project has a ShinyApp available online. Link: Capstone Project
Note: The app might be slow to start if the cloud infrastructure has put it into sleep state.

Tabular Playground Series - Mar 2021
The Tabular Playground Series is a set of competition on the Kaggle platform. On a monthly basis, a new challenge is proposed to the community.
Rank: Top 43% (633rd of 1495)
Link: Tabular Playground Series - Mar 2021

Housing Project
The housing transaction data are provided by the French government through an open source license. This is an example of an interactive map with the last transactions and per square meter price.

Android Sensors
Goal of this personal project is to build a dataset using a custom android app that captures the smartphone sensors (gyroscope, accelerometer, .?) during different human activities (walk, run, bike...).
Next step will be to build a machine learning model that detects the activity through the smartphone app.

Espresso or Longblack?
This project deals with how to handle a small image dataset with transfer learning.
The dataset and notebook are available on Kaggle.
Link: Espresso or Longblack?