Artificial intelligence-driven tool for spectral analysis: identifying pesticide contamination in bees from reflectance profiling

This dataset is associated with the scientific study titled "Artificial intelligence-driven tool for spectral analysis: identifying pesticide contamination in bees from reflectance profiling", which investigates the use of hyperspectral reflectance data and machine learning models to identify pesticide contamination in bee carcasses. This dataset can be used as an example on the MACSS platform (Machine Learning Applied to Chemometric Spectral Analysis) (https://github.com/bernardesrodrigoc/MACSS). MACSS was developed to facilitate the identification and to train models for other samples . The application is also available online as a web app at https://macsstat.shinyapps.io/macss. The dataset is available and consists of 3 primary files: data.txt: Format: DataFrame Content: Each row represents a sample of bee carcass reflectance data. Columns: The first column, labeled 'class', contains the categorical target labels indicating the type of pesticide contamination (dimethoate, fipronil, imidacloprid, glyphosate) or absence thereof. Subsequent columns represent numeric features corresponding to hyperspectral reflectance values at various wavelengths. Usage: This file is used for training machine learning models to predict pesticide contamination based on the reflectance profiles of bee samples. data_no_labels.txt: Format: DataFrame Content: Contains the same reflectance data as in data.txt but without the 'class' labels. Usage: This file is intended for simulating the classification of new samples where the contamination type is unknown, allowing for the evaluation and application of trained machine learning models in predicting contamination classes. complementary_data.txt: Format: DataFrame Content: Contains the same structure as reflectance data as in data.txt, but considering pesticide mixtures and varying contamination levels. The first column, labeled 'class', contains the categorical target labels indicating the type of pesticide contamination (fipronil, glyphosate in two different doses (22.2 µg a.i. bee-1 and 2.22 µg a.i. bee-1), and mix of fipronil + glyphosate) or absence thereof. Context and Application: The dataset is for developing and testing AI models that can classify bee contamination. The hyperspectral reflectance data, spanning hundreds of spectral bands, is utilized to train models capable of discerning pesticide contamination in different bee species, including <i>Apis mellifera, Melipona mondury, and Partamona helleri</i>.