1. Description: The BGS dataset was created by the British Geological Survey and describes geological measurements in Great Britain (http://data.bgs.ac.uk/). It was first used in~\cite{deVries2013 to predict the lithogenesis property of named rock units. The dataset contains around 150 named rock units with a lithogenesis, from which we used the two largest classes. 2. ML taks: classificaiton 3. Number of instances: 146 4. Original source: BGS 5. Linked to: BGS 6. Target variables: -"label_lithogenesis" (classification) -label_theme" (classification) 7. Stratified data split (training/test): -label_lithogenesis: TrainingSet(lith).tsv (80%) and TestSet(lith).tsv (20%) -label_theme: TrainingSet(theme).tsv (80%) and TestSet(theme).tsv (20%) 8. labels distribution - label_lithogenesis: label = GLACI =53; label = FLUV =93 - label_theme: label = BEDRCK =857; label = SPRFCL = 118