SNIC
SUPR
SNIC SUPR
Evaluating the FEST implementation of Random Forest for QSAR
Dnr:

SNIC 2017/7-241

Type:

SNAC Small

Principal Investigator:

Jonathan Alvarsson

Affiliation:

Uppsala universitet

Start Date:

2017-11-24

End Date:

2018-12-01

Primary Classification:

30101: Pharmaceutical Sciences

Webpage:

Allocation

Abstract

FEST (Fast Ensambles of Sparse Trees) is a C implementation of decision trees (on of them being the random forest algorithm which previously has been shown to be a good algorithm in this field) supporting sparse data from 2008. The fact that this implementation suports sparse data makes it very interesting since it means that it for our very sparse setups will need less memory at run time and bigger problems should be possible to study. We intend to evaluate the software and see if it performs well on various QSAR datasets.