Data Mining: Practical Machine Learning Tools and Techniques, Second Edition

selects three. In the event of a tie, we choose the rule with the greater coverage,

giving the final rule:

If astigmatism = yes and tear production rate =normal

and spectacle prescription = myope then recommendation = hard

This is indeed one of the rules given for the contact lens problem. But it only

covers three of the four hardrecommendations. So we delete these three from

the set of instances and start again, looking for another rule of the form:

If? then recommendation =hard

Following the same process, we will eventually find that age=youngis the best

choice for the first term. Its coverage is seven; the reason for the seven is that 3

instances have been removed from the original set, leaving 21 instances alto-

gether. The best choice for the second term is astigmatism=yes,selecting 1/3

(actually, this is a tie);tear production rate=normalis the best for the third,

selecting 1/1.

If age =young and astigmatism =yes and

tear production rate =normal then recommendation = hard

This rule actually covers three of the original set of instances, two of which are

covered by the previous rule—but that’s all right because the recommendation

is the same for each rule.

Now that all the hard-lens cases are covered, the next step is to proceed with

the soft-lens ones in just the same way. Finally, rules are generated for the none

case—unless we are seeking a rule set with a default rule, in which case explicit

rules for the final outcome are unnecessary.

What we have just described is the PRISM method for constructing rules. It

generates only correct or “perfect” rules. It measures the success of a rule by the

accuracy formula p/t.Any rule with accuracy less than 100% is “incorrect” in

110 CHAPTER 4| ALGORITHMS: THE BASIC METHODS

Table 4.9 Part of the contact lens data for which astigmatism=yesand tear

production rate=normal.

Age Spectacle Astigmatism Tear production Recommended

prescription rate lenses

young myope yes normal hard

young hypermetrope yes normal hard

pre-presbyopic myope yes normal hard

pre-presbyopic hypermetrope yes normal none

presbyopic myope yes normal hard

presbyopic hypermetrope yes normal none