Rustrees is an efficient decision tree and random forest library written in Rust with Python bindings. It aims to provide speed comparable to Sklearn with the reliability and performance of Rust.
- 🏎️ Speed: As fast as Sklearn on average.
- 🔗 Python Bindings: Effortless integration with Python.
- 🔒 Type Safety: Benefit from Rust's strong type system.
pip install rustrees
from sklearn.metrics import accuracy_score
from sklearn import datasets
import rustrees.decision_tree as rt_dt
df = datasets.load_breast_cancer()
model = rt_dt.DecisionTreeClassifier(max_depth=5).fit(df["data"], df["target"])
acc = accuracy_score(df["target"], model.predict(df["data"]))
print("accuracy", acc)
cargo add rustrees
use rustrees::{DecisionTree, Dataset, r2};
let dataset = Dataset::read_csv("iris.csv", ",");
let dt = DecisionTree::train_reg(
&dataset,
5, // max_depth
Some(1), // min_samples_leaf
Some(42), // random_state
);
let pred = dt.predict(&dataset);
println!("r2 score: {}", r2(&dataset.target_vector, &pred));
First, create a virtualenv (this just needs to be done once):
python -m venv .env
Then, activate the virtualenv (needs to be done every time):
source .env/bin/activate
Now, install the requirements (just needs to be done once):
pip install -r requirements.txt
Finally, install the Python library at the local virtual environment with the following command (needs to be done every time you change the Rust code):
maturin develop --release
Now, you can import the library rustrees
in Python. This can be done also from Jupyter notebooks. To do so, run the following command:
jupyter notebook
And then import the library in the notebook:
import rustrees.decision_tree as rt_dt