New RapidMiner user seeking advice

KMC_PhD

Hello

I am greatly hoping that this isn't a silly question, I am just getting started with RapidMiner.

I have a a spreadsheet which contains 24 rows of information on a student engagement with a virtual learning environment. I want to be able to classify students into certain ability groups based on the data on the spreadsheet, which is complete for each user.

My question is, can i explicitly state i.e. if the student spends more than a certain about of time on the virtual learning environment, accesses a forum, gets more that 60% in a quiz, then classify the student as learner type A.. if so what is the best way of going about this i.e. decision tress / association rules.

Any help is appreciated
 

MariusHelf

Hi,

first of all, 24 rows, i.e. 24 examples, is a very small learning base for any kind of machine learning algorithm, especially if you have a lot of columns, i.e. attributes. You probably won't be able to automatically train any (good) decision tree based on a data set that small.

To create a new attribute called Type which is set to certain values based on manual rules, you should have a look at the Generate Attributes operator. There you are able to specify e.g. rules with the syntax

if(quiz_value > 60 && time_spent > 10, "A", "B")

which will create a new attribute and set it to A if the condition matches and to B otherwise.

For an introduction to the general concepts of RapidMiner I'd like to direct you to the video tutorials on our website.

Best regards,
Marius

KMC_PhD

Thank you very much for the reply

As a correction I am using 24 attributes, currently for testing with aprox 200 rows of data, which will increase to apox 1000 rows..

Would your solution still be advisable with the increase in data? 

MariusHelf

If you want to stick with manual rules, yes. 1000 users should however also be a good basis for automated approaches - you could try a clustering algorithm and then try to describe the clusters to see according to which rules an algorithm would classify your users.

Best regards,
Marius

KMC_PhD

Thanks for the further info.. i think i will stick to the manual rules for now and see how i get on