PA196: Project summary
Dataset: Wine Quality (http://archive.ics.uci.edu/ml/datasets/Wine+Quality)
• 12 attributes
• unbalanced (much more average samples than extremes)
• could be used for regression or classification
Goal
• create / train reasonable classifier, that will predict wine quality (last attribute, ranging
form 1 to 10) from remaining 11 attributes
Planned approach
• balance / normalize dataset if necessary (e.g. if method takes to count quantity of
observed values)
• perform feature selection (probably not all attributes are relevant)
• evaluate multiple methods for classification
• compare with results published in original paper
• Multiplayer Perceptron
• SVM
• try some unused methods, i.e.:
• kNN
• LDA
• ???
Richard Trebichavský, 27. 10. 2015