Context: I'm currently writing my bachelor thesis and have decided to focus on what factors that contribute to students that have risky alcohol habits at my university. I am planing on doing a big survey to gather data about the students habits.

First Possible Approach: Since the classification problem is alcohol consumption I am having a slight issue in phrasing the question and its options. A similar study worked with a dataset based on educational data mining that used two measures Daily=daily alcohol consumption and Weekly=weekly alcohol consumption. The measures were 1 - very low to 5 - very high. Then they calculated the consumption as such:

(Weekly * 2 + Daily + 5) / 7

If the value was > 3 then he/she was classified as big drinker and if the value was < 3 he/she was not classified as a big drinker.

Second Possible Approach: However each year my university sends out a big survey to gather data about how much alcohol our students drink. They define a risky alcohol consumption as such:

  1. If you drink less than once a month then you have a low risk.
  2. If you drink 1-3 times a month then it means an increased risk.
  3. If you drink 1 time a week or often then that means you're in the risk zone.

Problem: What are you thoughts on the matter? I am not an data mining expert and that's why I am turning to you guys. Is it necessary for a binary classification as the similar study with a delicate matter as alcohol consumption? Or is perhaps 3-5 options as a measure more suitable?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.