I'm teaching myself Data mining and now struggling from this problem.

According to this table(http://slidewiki.org/slide/22352), if the attributes are Asymmetric, dissimilarity is calculated without t (Values which are 0 for both)

Let me give an example.

There are 2 objects, called Products.

They have 2 attributes, COLOR and SIZE.

  • COLOR is either RED(1) and BLUE(0) (They are binary).
  • SIZE is either BIG(1) or SMALL(0) (They are binary too).

When there are RED-BIG(1-1) & RED-SMALL(1-0) objects, d is 1/2, regardless of if the attributes are symmetric or asymmetric.

However, if they are BLUE-SMALL(0-0) and BLUE/BIG(0-1), d is 1/2 for symmetric BUT 1/1 for asymmetric.

How does the difference come from? In both cases, there is only one difference, BUT dissimilarity isn't the same.


1 Answer 1


I know it is the old post, but I also search for the answer to this question. and here what I got:

first of all, binary asymmetric attributes only care the 1 values. for example in measuring the distance of the patient based on their symptom, 1 means the symptom exists and 0 mean it does not exist. Or another example in measuring the distance between 2 documents based on the occurrence of words/term (most of these cases not using binary, but still asymmetric), 0 means the term does not exist in the document.

in those 2 cases, the number of symptom or term can be very huge and may cause a lot of 0 values appear in 1 records. This is because one disease may only have a small number of symptoms or one document may only contain a small number of words/term. If all of these 0 values are used to calculate the distance between two records, the similarity can be huge. example

record 1 : 1 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

record 2 : 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

assymetric: d(1,2) = 3/4 --> similarity(1,2) = 1-3/4 = 1/4

IF we consider it as symmertic: d(1,2) = 3/20 --> similarity(1,2) = 1-3/20 = 17/20 (large similarity)

that is why the attributes with the same 0 values between two records are not included in the distance calculation.


Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.