About Classifion

Take a tour

Inside the box

Brief theory of ...

List of references

Glossary of terms

Download & Order

Spectrino website

Sicyon website

 Written by and
Teodor Krastev

  • Statistical classification (or classification) - a statistical procedure in which individual items are placed into groups based on quantitative information on one or more characteristics inherent in the items (referred to as traits, variables, characters, etc) and based on a training set of previously labeled items.

  • One class classifier - classification without negative samples. The training does not take into account any other, but the the samples from that particular class. Positive - better general precision, easy on calculations and convenient for automation (as training creation and unknown classification); negative - the visual control is not as intuitive as in multiclass classifier.

  • Spec - spectrum usually mass-spectrum, but could be well optical or other. Correspond to a XY ASCII file.

  • Spec-group (or group) - set of spectra with something common in between. Usually separate measurements of the same sample. A group corresponds to one class/cluster in discriminant or cluster analysis.

  • Spectrino - Spectrino software stand-alone application and add-on to R  and it is identical to spectral data part of Classifion. It organizes the data, has visualization tools and contains pre-processing options.

  • Module - a functional unit, user-accessible as module page (bottom half of the workplace) or thru COM-server interface.

  • Type I error, or false negative - misclassification of sample which belong to a class, not to belong to that class.

  • Type II error, or false positive - misclassification of sample which does not belong to a class, to belong to that class.

  • Principal components  (or loadings, or factors) - orthogonal vectors to which the spectral data have been decomposed.

  • Cross-validation - consecutive excluding of one sample from a set and test that sample as unknown against the rest of them

  • Self-prediction - test a spec from the training set against the whole training set.

  • Cross-classification - same as self-prediction, but for groups instead of specs.