Distinguishing two datasets I have two datasets from some Web store (like Amazon). Datasets have one and the same structure. Each record in these datasets has the following attributes:
ID - user ID
ProductCnt - count of products user bought
DepartmentCnt - count of departments user was shopping  
PayTotal - sum of payments made by user   
PayCnt - count of payments
MaxPay - maximum payment

The first dataset is a collection of records related to different users randomly selected. The second dataset is a collection of records related only to users who also clicked on a particular advertisement located on the same page as the product they where shopping for. 
Problem: Find dependencies that distinguish users in second dataset from users in the first dataset.
To solve this I would calculate statistical parameters such as mean, expected value and standard deviation for all parameters in both datasets and compare them.
Any other ideas how to find characteristic features distinguishing these datasets? 
I am new to this kind of problems, so please bear with me, and also let me know if my question makes no sense at all!
Thanks!
 A: It sounds like you are facing a semi-supervised classification problem. Based on your problem statement you have a positive set (second data set) and an unlabeled set (first data set). The first data set is unlabeled because it can contain both users that clicked the ad and users that didn't.
This kind of problem is a twist on traditional supervised classification. It poses a bunch of additional challenges, the main one being model selection, since the usual measures like accuracy are not useful.
If you look up techniques to learn from positive and unlabeled data (often called PU learning), you will probably find several useful ideas. The key idea that is common in most PU learning approaches is to weigh positive training instances (significantly) higher than unlabeled ones.
If you want to determine which features are relevant, you may consider approaches like semi-supervised logistic regression. I consider Learning Classiﬁers from
Only Positive and Unlabeled Data by Elkan and Noto to be a very good reference on the subject.
