Research on Distinguishing Biological Species by Data Model and Linear Discriminant Analysis

Abstract

This article explores the classification of lizards based on their distinct pholidosis and morphological characteristics using various data attributes. The authors aim to construct a classification model that takes advantage of data attributes for both simplicity and accuracy. Additionally, the article aims to propose an adaptive model that provides recommendations according to the precision requirements of biologists and the computational environment, enhancing the model's applicability. The authors employ Fisher's and Bayesian methods from linear discriminant analysis for classification, leveraging the linear structure to ensure the model's simplicity. A novel aspect of this work is the development of a discriminative power index for variables. This index prioritizes variables with strong discriminative abilities, thus simplifying computations and improving efficiency. The results align with those obtained through exhaustive searches for optimal solutions. Furthermore, the constructed model offers classification criteria and prediction accuracy under different variable combinations, enabling biologists to adjust variables based on accuracy needs and computational constraints. This functionality enhances the model's suitability for various real-world research scenarios.