This is part 7 of a tour through PEAXACT - Software for Quantitative Spectroscopy from S-PACT. It's now time to look into the Classification Model and how it can be used to identify categorical features from unknown samples.
Classification Model
A Classification Model uses the spectra to discriminate classes of samples. All methods relate back to the spectral similarity between the samples. In the training step, the model learns how similar samples inside a class are, and where unambiguities between classes occur.
Training a Classification Model requires categorical features, e.g., material IDs, to form classes of samples with identical values. Pretreatment emphasizes the spectral characteristics of each class and helps to distinguish between them. In an ideal case, all given classes can be perfectly separated and identified. The Cluster Plot in the Data Inspector is a helpful preview, but final assessment of the most suitable Classification Model is typically done with the Confusion Matrix.
Identification Analysis
When analysing unknown samples, the Classification Model assigns them to the class with highest similarity. The Confusion Matrix Plot (see above) is the quickest way to assess the performance of the Classification Model. In addition, the assignment of a sample to a class is accompanied by a Class Probability ("p-value"), visible, e.g., in the Report Table.
Classification Models can be deployed to field instruments for automated real-time analysis - just like any other PEAXACT model. But first let's take a look into one last but very powerful part of the model.