Scanning hyperspace: how to tune machine learning models

The distribution of the outcome variable quality is a bit funky – the values are mostly 5 and 6 (how would you check this?). This could get a bit irksome later on, so go ahead and recode the quality scores into something more convenient. One idea would be to label wines as being either high quality (e.g. if their score is 6 or higher) or low quality (if the score is 6 or lower).