Train SizeTest SizeLength Number of ClassesNumber of DimensionsType
261 2631751 43SPECTRO
Donated By: James Large, UEA
Description: EthanolConcentration is a dataset of raw spectra of water-and-ethanol solutions in 44 distinct, real whisky bottles~\cite{large2018detecting}. The concentrations of ethanol are 35\%, 38\%, 40\%, and 45\%. The minimum legal alcohol limit for Scotch Whisky is 40\%, and many whiskies do maintain this alcohol concentration. Producers are required to ensure that the contents of their spirits contain alcohol concentrations that are tightly bound to what is reported on the labelling. The classification problem is to determine the alcohol concentration of a sample contained within an arbitrary bottle. The data has been arranged such that each instance is made up of three repeat readings of the same bottle and batch of solution. Three solutions of each concentration (batches) were produced, and each bottle+batch combination measured three times. Each reading is comprised of the bottle being picked up, placed between the light source and spectroscope, and spectra saved. The spectra are recorded over the maximum wavelength range of the single StellarNet BLACKComet-SR spectrometer used (226nm to 1101.5nm with a sampling frequency of 0.5nm), over a one second integration time. Except for avoiding labelling, embossing, and seams on the bottle, no special attempts were made to obtain the cleanest reading for each individual bottle, nor to precisely replicate the exact path through the bottle for each repeat reading. This is to replicate potential future conditions of an operative performing mass-screening of a batch of suspect spirits. @inproceedings{large2018detecting, title={Detecting Forged Alcohol Non-invasively Through Vibrational Spectroscopy and Machine Learning}, author={Large, James and Kemsley, E Kate and Wellner, Nikolaus and Goodall, Ian and Bagnall, Anthony}, booktitle={Pacific-Asia Conference on Knowledge Discovery and Data Mining}, pages={298--309}, year={2018}, organization={Springer} }
Download this dataset
Dataset Image

Best Algorithm:
Best Accuracy: