Welcome to my machine-learning repo

Kaggle's Porto Seguro safe driver competition:

• Complete model

Solution to this classification problem using different techniques : logistic regression, XGBoost classifier and DNN (tensorflow). Features selection is performed with ExtraTreesClassifier, and PCA can be used if the data is binarized (leading to too many features).

Data preprocessing is done with a specific class.

Kaggle's Titanic competition:

• Random forest using GraphLab

• Basic decision tree using GraphLab

• Home-made k-NN model

• Multiple models with scikit-learn

This model combines algorithms such as : RandomForestClassifier, AdaBoosClassifier, DecisionTreeClassifier, SVM, KNN. For the final prediction EnsembleVotingClassifier is used. Best score is 0.79904

Kaggle's House price competition:

• Jupyter notebook, multiple models with scikit-learn

This model combines several regression techniques LASSO, Elastic Net, Gradient Boosting, AdaBoost, XGBoost. Best score is **~ 0.1250**

• Using MLBox

MLBox is a python automated machine-learning library, written by Axel de Romblay

For more details see GitHub Flavored Markdown.