Random Forests

The goal of this project is to implement in Python a random forest for classification, regression and outlier detection. The focus, however, is on two aspects:

• object-oriented design, according to OO principles and patterns
• good programming practices: logging, comments and coding style

This project was the practicum of the Object Oriented Programming course of the Matemàtiques Computacionals i Anàlisi de Dades (Mad-CAD) degree at Universitat Autònoma de Barcelona, 2021-22.

Requirements

• Python 3.x
• Numpy
• Matplotlib

Usage

Follows scikit-learn style

rf = RandomForestClassifier(
        max_depth,    # of each decision tree
        min_size,     # of a node to make it a leave
        ratio_sample, # to get the size of the dataset when looking for the best split
        n_trees,      # number of trees in the forest
        n_features,   # number of features (randomly selected) to consider when looking for the best split 
        criterion     # either Gini or Entropy              
)
rf.fit(Xtrain, ytrain)
ypred = rf.predict(Xtest)
# and same for regression
rf = RandomForestRegressor(...)

Design

(put here your final design)

Authors

This is team work with Pere Garriga and Genis Soler.