Packages and Releases | |
---|---|
Build Status | |
Test Coverage | |
Other Information |
A fast-and-frugal-tree classifier based on Python's scikit learn.
Fast-and-frugal trees are classification trees that are especially useful for making decisions under uncertainty. Due their simplicity and transparency they are very robust against noise and errors in data. They are one of the heuristics proposed by Gerd Gigerenzer in Fast and Frugal Heuristics in Medical Decision. This particular implementation is based on on the R package FFTrees, developed by Phillips, Neth, Woike and Grassmaier.
You can install fasttrees using
pip install fasttrees
Below we provide a qick first start example with fast-and-frugal trees. We use the popular iris flower data set (also known as the Fisher's Iris data set), split it into a train and test data set, and fit a fast-and-frugal tree classifier on the training data set. Finally, we get the score on the test data set.
from sklearn import datasets, model_selection
from fasttrees import FastFrugalTreeClassifier
# Load data set
iris_dict = datasets.load_iris(as_frame=True)
# Load training data, preprocess it by transforming y into a binary classification problem, and
# split into train and test data set
X_iris, y_iris = iris_dict['data'], iris_dict['target']
y_iris = y_iris.apply(lambda entry: entry in [0, 1]).astype(bool)
X_train_iris, X_test_iris, y_train_iris, y_test_iris = model_selection.train_test_split(
X_iris, y_iris, test_size=0.4, random_state=42)
# Fit and test fitted tree
fftc = FastFrugalTreeClassifier()
fftc.fit(X_train_iris, y_train_iris)
fftc.score(X_test_iris, y_test_iris)