On the previous page basic tests...
The further tests on implemented forest algorithms. All of these graphs were made using following databases:
- Car database
- Balance-scale database
- Ionosphere database
For more informations about used databases read the documentation or basic tests.
There were used following algorithms:
- Random forest (RI)
- Det. forest
- Forest RC
On each data set test function where run three times and the results were combined.
Test set errors:
Test set errors as a function of number of trees in forest.
Train set errors:
Train set errors as a function of number of trees in forest.
Increases in error rates due to noise.
The rate of instances classified to different class using forests grown with the same input parametres.