kushvarma / dm_arf

Data Mining Lab SS2020 Uni-Hannover. Adaptive Random Forest with Resampling

---

---

Adaptive Random Forest with Resampling for Imbalanced data Stream

Paper by: Luis Eduardo Boiko Ferreira, Heitor Murilo Gomes, Albert Bifet, Luiz S. Oliveira.

[Paper URL 1](https://ieeexplore.ieee.org/document/8852027)

[Paper URL 2](https://www.researchgate.net/publication/336152223_Adaptive_Random_Forests_with_Resampling_for_Imbalanced_data_Streams)

How to run?

Adaptive Random Forest with Resampling is not yet available in MOA. It has to be build. Use this Repo: https://github.com/kushvarma/moa.git and switch to branch arf_re to get the code. These instruction is for Intellij IDEA

• One you have the code, switch to branch arf_re
• Click on Edit Configuration on Top right on window
• Click on Plus on left of window, then select Application.
• After that fill following details:

• Click OK
• Press Play button next to MOA GUI on top RIght of Intellij
• It will open MOA GUI application.
  

Dataset

The data sets are available in the repo. There are two types of dataset, ARFF for MOA and csv to be used for scikit-multiflow. For scikit-multiflow, the dataset need to be cleaned and modified to run the experiment.

scikit-multiflow with Adaptive Random Forest with Resampling

We are also working on porting ARF_RE to python. The source code is available on https://github.com/kushvarma/scikit-multiflow.git and branch dm_arf.

Results from MOA and comparision to the PAPER

The current result is available in Result folder. Comparing result from the Paper.
All the test were run on Core i5 8400/ 32GB RAM machine.