Ranger is run from the command line and data are exchanged via text files. This is quick and dirty, not a clean OCaml interface to ranger. The oranger_rfr program allows to train/test a random-forests regressor model.
$ oranger_rfr usage: oranger_rfr [-p ]: proportion of the (randomized) dataset used to train (default=0.80) [-np ]: max number of processes (default=1) [-n ]: |RF|; default=100 [--mtry ]: proportion of randomly selected features to use at each split (default=(sqrt(|features|))/|features|) [--scan-mtry]: scan for best mtry in [0.001,0.002,0.005,...,1.0] (incompatible with --mtry) [--mtry-range ]: mtrys to test e.g. "0.001,0.002,0.005" [-o ]: output scores to file [--train <train.txt>]: training set (overrides -p) [--valid <valid.txt>]: validation set (overrides -p) [--test <test.txt>]: test set (overrides -p) [--NxCV ]: number of folds of cross validation [--seed : fix random seed] [--no-regr-plot]: turn OFF regression plot [--rec-plot]: turn ON REC curve [--y-rand]: turn ON Y-randomization [-s ]: save model to file [-l ]: load model from file [--max-feat ]: max feature id. (cf. end of encoding dict) [-v]: verbose/debug mode [-h|--help]: show this help message
Published: 13 Apr 2022
Thin OCaml wrapper for the ranger (C++) random forests implementation. Ranger is run from the command line and data are exchanged via text files.
ranger is here: https://github.com/imbs-hl/ranger