# One-Class Genetic Algorithm for Authentication Analysis of Spectrochemical Data

## TL;DR

This paper introduces a new method combining genetic algorithms with one-class classifiers to improve disease diagnosis using spectrochemical data.

## Contribution

A novel one-class genetic algorithm (OGA) is proposed for variable selection in one-class classification of clinical spectrochemical data.

## Key findings

- The OGA-PRM-OCPLS model achieved 100% sensitivity for COVID-19 and endometriosis classification.
- The OGA-DD-SIMCA model achieved 100% sensitivity for dengue classification.
- OGA-selected variables can be linked to biomarkers for disease diagnosis.

## Abstract

One-class classifier (OCC) models are widely applied
to solve classification
problems where control or class modeling from a target class is necessary.
In this study, OCC models such as Data Driven Soft Independent Modeling
of Class Analogy (DD-SIMCA) and One-Class Partial Least Squares (OCPLS)
were associated with a new variable selection strategy, the one-class
genetic algorithm (OGA), for classification analyses in three clinical
applications: COVID-19, endometriosis, and dengue samples. DD-SIMCA
was implemented in a rigorous approach, using α = 0.05, while
OCPLS was performed with partial robust M-regression (PRM). For the
three cases, a better classification performance was obtained using
the OGA associated with the OCC model. The performance of the OGA-PRM-OCPLS
showed better results for both COVID-19 and endometriosis cases when
compared to DD-SIMCA, with a classification sensitivity of 100%. However,
the best results for dengue classification were obtained by using
the OGA-DD-SIMCA model (sensitivity = 100%). The selected variables
obtained by the OGA can be used to relate this information to biomarkers
capable of distinguishing between case and control groups. These findings
have the potential to improve some disease diagnosis using chemometrics
for the development of rapid, low-cost, and minimally invasive screening
methodologies.