Classifiers

Fast and reliable implementations of various classifiers can be provided, featuring, amongst others, the following properties:

• threshold-based decision rules, exportable as boolean expressions,
• feature ranking and reduction,
• optimal consideration of NA-values (not available features),
• linear and nonlinear decision boundaries,
• a parameter controlling the complexity of trained classifiers.

The computational burden can be considerably decreased while increasing the overall recognition rate by resorting to the optimal selection of describing features!

All solutions are embedded in a powerfull C++ library:

• the training state of a classifier can be saved and imported,
• some classifiers are parallelized on CPUs.