An proof-of-concept implementation of local linear forest. The implementaion followed the paper by Rina F. et al.
Python 3.7, two packages were used: sklearn 0.19.2, and numpy 1.15.4. It is built on top of the basic tree structure implemented in sklearn, the efficiency is not optimized in this POC implementation.
The local linear forest shows some improvement on the "step function" in the ordinary random forest. See notebook
<
To install, clone this repo to local disk and type
pip install -e .
The usage is the same as sklearn RandomForestRegressor
from LLF.LocalLinearForest import LocalLinearForest
llf_1 = LocalLinearForest(n_estimators=100, max_depth=5)
llf_1.fit(X, y)
llf_1.predict(X_test)