Tuned Data Mining in R
| Package: | TDMR |
| Type: | Package |
| Version: | 2.2 |
| Date: | 01.03.2020 |
| License: | GPL (>= 2) |
| LazyLoad: | yes |
TDMR is a package for tuned data mining (predictive analytics, i.e. classification and regression). Its main features are:
1) A variety of tuners, with special emphasis on SPOT (a well-known R package for parameter tuning), but also CMA-ES
(package rCMA-package) and other tuning algorithms.
2) Tuning of preprocessing parameters and model building parameters simultaneously. Preprocessing
often includes feature generation.
3) Support for multiple tuning experiments (different settings, repetitions with different resamplings, ...).
4) Easy parallelization of those experiments with the help of R package parallel.
5) Extensibility: New tuning parameters, new preprocessing tools, model builders and even new tuners can be added easily.
The main entry point functions are tdmClassifyLoop, tdmRegressLoop,
tdmTuneIt, and tdmBigLoop.
See tdmOptsDefaultsSet and tdmDefaultsFill for an overview of adjustable TDMR-parameters.
http://lwibs01.gm.fh-koeln.de/blogs/ciop/research/tuned-data-mining/