Ing. Marcel Jiřina, DrSc.

Vědecký pracovník


Project FRF - Friendly Random Forest

The Random Forest is a method and also a program for data clustering and classification. For basic reference see Breiman, L.: Random Forests, Machine Learning Vol. 45, No. 1, pp. 5-32. (2001). Especially in classification the Random Forests method appears generally to be the best approach perhaps among all others up to now.

Motivation

We found the original Random Forests program rather difficult to use. The necessity to set up (change) parameters in the source text of the FORTRAN program is rather effective for tuning and experimentation. For user who has no knowledge about Fortran 77 programming language and about peculiarities of the use of a compiler, it may be a stressful task. This feeling may result in the use another classifier that is friendlier to use.

Friendly Random Forest

In our modification in Fortran 90 the program is much friendlier than original version keeping the program core untouched. Binaries for Windows and Linux as well as Fortran source code available in the package [here]. Information for different tasks is passed with help of arguments. In the report [here] after a brief description of our modification of RandForest the detailed manual follows. The next Chapter describes testing data corpora used and gives results of the Random Forest program.

References and publications

  • Basic:
    • Breiman, L.: Random Forests, Machine Learning Vol. 45, No. 1, pp. 5-32. (2001)
  • Our modification:
    • Jiřina, M., Jiřina, M., jr.: Testing Random Forest for Unix and Windows. Technical Report No. V-1075 , Institute of Computer Science AS CR, Prague, Czech Republic. (2010)
    • Marcel Jirina, M. Said Krayem, and Marcel Jirina, Jr. Adapting GNU Random Forest program for Unix and Windows. 11th International Conference of Numerical Analysis and Applied Mathematics 2013 AIP Conf. Proc. 1558, pp. 337-340 (2013); doi: 10.1063/1.4825492 AIP Publishing LLC 978-0-7354-1184-5 (2013)

Software

This software is freeware with no warranty and for non-commercial use only.

Download: FRF Package