Stochastic Local Search Decision Tree
This repository is for my first scientific initiation project.
Decision tree is a predictive modelling aproach used in machine learning, data mining and statistics. In the decision tree each internal node represents a test on a feature and each terminal (or leaf) node represents a class label. Oblique Decision Tree is a variation of traditional decision trees, which allows multivariate tests in its internal nodes in the form of a combination of the features.
Our research, SLSDT, is a method for induction oblique decision trees using a stochastic local search method called Late Acceptance Hill-Climbing (LAHC) to try to find the best combination of features in each internal node.
This project also provides a utility to read csv files and convert to the format accepted by the SLSDT method. Moreover provides also a utility to load datasets included in the package.
- Install
pip3 install slsdt- read_csv
from slsdt.reader_csv import read_csv
X, y = read_csv("iris.csv", "class")- load_dataset
from slsdt.datasets import load_dataset
X, y = load_dataset("iris")- slsdt
from slsdt.slsdt import SLSDT
# split train and test data
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.33, random_state=42)
clf = SLSDT()
clf.fit(X_train, y_train)
results = clf.predict(X_test)
print(f"Accuracy: {sum(results == y_test) / len(y_test)}")from sklearn import datasets
from slsdt.slsdt import SLSDT
iris = datasets.load_iris()
X = iris.data[:, :2] # we only take the sepal width and sepal length features.
y = iris.target
mark = y != 2
# we only take the 0 (Iris-setosa) and 1 (Iris-versicolor) class labels
X = X[mark]
y = y[mark]
clf = SLSDT()
clf.fit(X, y)
clf.print_tree()
result = clf.predict(X)
print(result)
print(result == y)
Plot with Matplotlib using the results obtained above.
- Leave the ⭐ if you liked the project
- Fork this project
- Cloner your fork:
git clone your-fork-url && cd slsdt - Create a branch with your features:
git checkout -b my-features - Commit your changes:
git commit -m 'feat: My new features' - Send the your branch:
git push origin my-features
This project is licensed under the EPL 2.0 License - see the LICENSE file for details.