Another important feature of YAILS during classification is the weighted flexible matching. This mechanism is two-folded. First it results from the ability of attributing a value to each selector of one rule. The goal of this value is to distinguish between the importance of the selectors in one rule towards the conclusion. The second issue is the fact that YAILS is capable of using a rule even if some of its conditions are not satisfied by the example. These two factors are mixed together as we will see thus given YAILS the ability of performing weighted flexible matching.
The calculation of each selector's value is done using the concept of entropy(H) of a complex. Each rule has its entropy. Given that value YAILS calculates the decrease in entropy that each selector caused when it was added to the rule. This calculation is done as follows :
Rule-s is the generalization of Rule obtained by removing s from Rule
With each rule having its selectors characterized by one value YAILS is capable of calculating the matching score (MS) of one example with respect to the rule. This matching score is calculated using the following formula:
where
Example :
Every time an example needs to be classified YAILS calculates its MS with respect to every rule belonging to the current FS. Notice that there is a user definable parameter which sets the minimal matching score for a rule conclusion to be considered as a possible classification.