Diachrysia classification

This repository contains the data, code, and results for “Reflectance spectroscopy and machine learning as a tool for the categorization of twin species based on the example of the Diachrysia genus” article.

Dataset

There are two datasets in the data directory - the first containing legislative information on moths (Legislative), the second containing the measured spectra (Spectra). The files are available in two formats - .csv and .xlsx.

Legislative file contains the following columns:

• ID - Individual identifier (species-number)
• Species - Diachrysia chrysitis or Diachrysia stenochrysis
• Sex - Male (♂) or Female (♀)
• Year_catch - year of the moth catch
• Day_catch - day of the year when the moth was caught
• Locality - place where the moth was caught
• UTM_code - zone in UTM coordinate system
• Longitude - east–west position in degrees
• Latitude - north–south position in degrees
• Feature_level - level of marking the morphological feature, where 1 means weak, 2 means strong, 3 means very strong

Spectra file contains the following columns:

• ID - Individual identifier (species-number)
• Species - Diachrysia chrysitis or Diachrysia stenochrysis
• Scale - part of the scale on which the spectrometer measurement was made (Glass or Brown)
• 400-2100 - spectral band number

Reproduction

1. Open the diachrysia-classification.Rproj project file in RStudio.
2. Run 01_randomforest.R to build classification models, assess the performance of classification at the general and individual level and determine the importance of spectral features.
3. Run 02_KS_test.R to determine importance of the spectral bands for species discrimination using Kolmogorov–Smirnov test.
4. Run 03_LDA_best_features.R to determine the most useful spectral bands for species classification using Linear Discriminant Analysis and D-statistic.
5. Run 04_LDA_combinations.R to determine the minimum set of spectral features to distinguish species with 100% accuracy.

Results

The code results were saved in results directory:

• ks-test.csv - importance of the spectral bands for species discrimination determined by D-statistic
• rf-importance.csv - average importance of the spectral features for classification in the random forest models