Target class feature model

An apparatus comprising:

• a non-transitory computer-readable medium containing instructions to direct a processor to:
• receive a composite of data features from each data sample of first data samples having a membership in a target class;
• receiving a composite of data features from each data sample of second data samples not having a membership in the target class;
• extract and quantify individual data features from each composite of the first data samples and the second data samples;
• iteratively apply different candidate weighting combinations to feature sets to create a different candidate model during each iteration;
• for each iteration, apply the different candidate model to the extracted and quantified individual features from each composite of the first data samples and the second data samples;
• compare a result of each different candidate model applied to the extracted and quantified individual features from each of the first data samples and the second data samples to a known membership status of each of the first data samples and the second data samples;
• and
• select the target class feature model representing the fingerprint of the target class based upon composites of individual features based upon the comparison, wherein the first data samples and the second data samples comprise SERS spectra data and wherein the feature sets comprise features based upon spectral peaks of the SERS spectra data.

A method comprising:

• sensing first data samples from a first set of different subjects having a membership in a target class;
• sensing second data samples from a second set of different subjects not having a membership in the target class, each of the first data samples and the second data samples comprising a composite of individual data features;
• extracting and quantifying the individual data features from each composite of the first data samples and the second data samples;
• and
• identifying sets of data features and associated weightings of a target class feature model based upon quantified values of the individual data features from each composite of the first data samples and the second data samples to create a target class feature model representing a fingerprint of the target class to determine membership status of a sample having an unknown membership status with respect to the target class, wherein the first data samples and the second data samples comprise SERS spectra data and wherein the sets of data features comprise features based upon spectral peaks of the SERS spectra data, wherein the first data samples and the second data samples comprise SERS spectra data of a subject taken with different Raman spectroscopy enhancement surfaces.