Overview

Streaming Decision Trees & Forests: exploring streaming options for decision trees and random forests.

The package includes two ensemble implementations (Stream Decision Forest and Cascade Stream Forest).

Based on scikit-learn fork.

Install

You can manually download the latest version of SDTF by cloning the repository:

git clone https://github.com/neurodata/SDTF
cd SDTF
python setup.py install

Or install the stable version through pip:

pip install sdtf

Package Requirements

The SDTF package requires a scikit-learn fork for the partial_fit functionality, which you can install manually:

git clone https://github.com/PSSF23/scikit-learn-stream
cd scikit-learn-stream
python setup.py install

The above local setup requires the following packages:

• cython

• numpy

• scipy

Relevant Repos

• online-ml/river

• scikit-garden/scikit-garden

• huawei-noah/streamDM

• soundcloud/spdt

Relevant Papers

• Very Fast Decision Tree

• Mondrian Forests

• Online Bagging and Boosting

• Leveraging Bagging for Evolving Data Streams

• Ensemble Learning for Data Stream Classification

• Adaptive Random Forests

• Streaming Random Forests

• Streaming Parallel Decision Tree

Benchmarks

• Benchmark Figures on Selected Tasks
• Benchmark Figures on OpenML-CC18
• Gaussian XOR Experiments

API

• API Reference