1. Data preprocessing and exploratory analysis

• [ ] deal with missing values and categorical variables (use one-hot encoding)
• [ ] check for correlated variables
• [ ] analyse distributions of attributes
• [ ] visualise the data using two or three principal components

2. Logistic Regression model

3. More advanced models

3.1 Random Forest

3.2. Support Vector Machine

3.3. Feed Forward Neural Network

For each model:

• standarize the dataset
• use the same cross-validation scheme
• inspect feature importance
• evaluate model with logloss, ROC-AUC and F1 classification metrics

Resources

• dataset
• paper
• Python Machine Learning by Sebastian Raschka
• interesting talks:
  • Machine Learning 101 by Kyle Kastner (+ GitHub repo)
  • Classification using Pandas and Scikit-Learn by Skipper Seabold (+ GitHub repo)
  • Machine Learning with Scikit-Learn by Jake VanderPlas (+ GitHub repo)
  • Neural Nets for Newbies by Melanie Warrick (+ GitHub repo)

Keywords:

• dimensionality reduction
• cross-validation
• supervised vs unsupervised learning
• regression vs classification
• confidence intervals and p-values

Tools

• conda
• jupyter notebook
• pandas
• sklearn
• seaborn