cancel
Showing results for 
Search instead for 
Did you mean: 

Matching/Merging- Threshold calculation

Former Member
0 Kudos

Hi, I am facing problem to define High/Low Threshold value. What is the exact approach to define it?

Because of wrong threshold value I am not getting proper count value.

Let me know as soon as possible...

Note: Plz don't send me any link to check documents. I am aware about all documents...

Thanks in Adv,

Ron

Accepted Solutions (1)

Accepted Solutions (1)

michael_theis
Active Contributor
0 Kudos

Hi Ron,

I think there is no general exact approach to define those values. It depends on your data and your matching strategies.

When defining the MDM Matching Rules you have to define scores for a successfull and unsuccessfull execution of the rule. A matching strategy consists of one ore more rules. This sums up the different scores of the rules and results in a minimum and maximum score. You can use the thresholds to define ranges (no match, low match, high match). This thresholds should be defined in regard to your matching rules.

An example:

Let's assume your matching strategy consists of only one rule. You use the "Equals" function which results in "full match" or "no match". You define a score of 100 for the successfull match. In this case, the only the high threshold of 100 would make sense as your matching strategy would result with 0 or 100 only.

A second example:

Your matching strategy consists of one rule, too. But you use now the "Token Equals" because the field you're matching may consist of several tokens. The success score is set to 100 again. Hence you've defined the "Token Equals" the resulting score may vary depending on the number of matched tokens. Now you can use the threshold as a "percentage". Assuming you always match 5 tokens against each other and you want to define that at least 4 tokes have to match the receive a "High" matching class, it would make sense to set the high threshold to a score of 80.

Kind regards

Michael

Answers (1)

Answers (1)

former_member189059
Active Contributor
0 Kudos

hi ron,

Threshold values indicate the score required to show up as a difference

if it is below the lower threshold, the comparison Class shows as 'NONE' and count is 0 (red color)

Once the lower threshold value is crossed, the Class is set to 'Low' and the count is increased to 1(blue color)

Once the higher threshold value is crossed, the Class is set to 'High', count remains at 1, but the color is Green

the score is the sum of the scores of all the rules for a particular strategy that you apply

(according to Success and Failure) parameters

Former Member
0 Kudos

Thanks Michael/Kris. It's helpful....

Kris: Just want to confirm one thing. I may be wrong.

Count is not depend on Low or High. It will not be 1 always. Actually count is the number of records which are duplicate corresponding to that particular record. By selecting Matches tab(First tab in M mode) we will get the number of record corresponding to selected record in "Records" pane.

Let me know Am I right or my understanding is wrong.

former_member189059
Active Contributor
0 Kudos

yes, you are right, 'count' shows the number of duplicates

Former Member
0 Kudos

Thanks a lot Kris/Micheal...

I am closing the thread..