Calculate weights for a similarity score

Question

Assume there is a dataset as follows:

dataA: {
  attribute1: x,
  attribute2: y,
  attribute3: z
}

I want to calculate the correlation between similar structured data (Eg: dataA, dataB, dataC ... )

And I have a similarity measure for each attribute of each dataset. (Eg: similarity of x with other values of attribute1 is 0.11, similarity of y with other values of attribute2 is 0.22, similarity of z with other values of attribute3 is 0.33)

I'm going to present the correlation score in a weighted average approach where a weight is defined for each attribute (Eg: weight of attribute1 is w1 etc.):

Score for dataA = { (0.11 x w1) + (0.22 x w2) + (0.33 x w3) } / {w1 + w2 + w3}

If I'm going to conduct an experiment to find the optimal weights, how can I do it?

UPDATE:

Can I conduct an experiment to check the probability of each attribute value to be changed and then use that value somehow?

Answer 1

分配问题或单纯形法怎么样？