Optimization of Natural Dyes' Non-contact Characterization and Interdisciplinary Application Using Ensemble Classifiers and Genetic Algorithms

Abstract

The reported evaluation of the problematic encountered by noncontact characterization of natural dyestuffs demonstrates that practically no one analytical method alone – destructive, less, or non – can warrant absolute reliability of the results. The main obstacles are due, in general, to our limited knowledge on reactivity of natural compounds, their sources, methods of proceeding and application; to the complexing role of the ambient/medium on the photo reactions, as well as to the actual possibilities of analytical techniques; to the lack of commonly accepted standards amongst different scientific traditions, etc. For to overwhelm these drawbacks and optimize accuracy of NDs characterization, we propose the integration of chemio-physical with computational assessment. The proposed approach consists in calculation the probability of a hypothesis comparing new results with already stored interdisciplinary data using the high performance ensemble techniques Random Forest and Random Subspace based on the genetic approach termed EV-Ensemble approach. Its role is to render more accurate the analytical results, to avoid sampling and simplify the identification analysis, to facilitate interdisciplinary applications of NDs in high technologies, and to share generated experience between the research community.

Keywords

• Natural Dyestuffs
• Knowledge Representation
• Ensemble Classifiers
• Random Forest. Random Subspace