Infrared spectral data of dried and roasted cocoa (Theobroma cacao L.) for calibrating classification models of cocoa varieties and predicting theobromine and caffeine content

The dataset compiles the mid-infrared spectra of dried and roasted cocoa beans (Theobroma cacao L.), as well as their theobromine and caffeine content. The Attenuated Total Reflectance-Fourier Transform Infrared (ATR-FTIR) spectroscopy technique was employed to spectrally characterize the dried and roasted cocoa samples, while high-performance liquid chromatography (HPLC) was used to quantify the theobromine and caffeine content. Based on the theobromine/caffeine ratio, the cocoa variety was determined.
This dataset provides valuable data for researchers, food scientists, and industry professionals to develop and calibrate robust multivariate statistical models, as well as machine learning and deep learning techniques, to classify different varieties of dried and roasted cocoa beans in real-time. Additionally, the infrared spectra could be used to develop precise models for simultaneously predicting both theobromine and caffeine content based on the spectral data. These calibrated models could significantly support decision-making processes within the cocoa industry, enabling real-time quality control and optimizing bean selection, blending, and product formulation processes. The dataset is organized in Excel sheets and structured according to experimental conditions and replicates, allowing for detailed exploration and analysis.