Partial least squares discriminant analysis (PLS-DA) can be used to identify metabolic differences between distinct classes by finding linear relationships between the spectral data and class variables, e.g. receptor status [ 25].
This grouping method is based on the ArcGIS implementation of the Fisher-Jenks algorithm and minimizes differences between data values within classes and maximises differences between classes [ 24, 25].
They are capable of performing complex non-linear mapping between the training data and their class membership, and they have been successfully used to analyze neural datasets [34] [36].
They define a set of expected and unexpected rules, their expectations being no relation between the data attributes and classes upon which stronger co-relations are investigated.
In this paper we compute mutual information between data points X and classes C. A large value of mutual information in this case means that we have much information about the class C given the observation X.
The matching algorithm automatically converts the data types between table and class objects using the reflection mechanism in the Java language, which provides the information of parameter names and types in the class definitions.
SVM is a kernel-based learner, which can find non-liner boundaries between data classes by using kernels.
Typically, the relationships between measured data and interesting properties or class memberships of the materials cannot be described by theory but require empirical models, that have to be derived from data originating from materials with known measurements and properties.
Based on our findings in the experiments described above, we propose a subspace clustering-based workflow (c.f. Fig. 11) to find relations between data records, dimensions, and associated class labels.
The distinction between label and class identifiers caters for changing metadata associated with the class without having to modify data that are already characterized with the class identifier.
