Practical Applications of Data Mining

3.3 separate-and-Conquer approaCh 97

After trimming off the third to the last term “age is presbyopic,” the P, T,

p, t values and the probability (Pr) measure are as follows:

T = 24, P = 15, t = 12, p = 3

Pr = 1.00

As seen in the example above, the original rule has the lowest probability

value. Therefore, we abandon the trimmed rules and keep the original one.

The same principle can be applied to rules 3 and 4. No conditional terms are

pruned off from these rules. Therefore, the rule list remains the same.

3.3.3 REP, IREP, RIPPER

The preceding procedure for generating rules leads to over-fitted rules

since it uses the same dataset to generate and evaluate the rules ([Pagallo

1990], [Brunk 1991], and [Cohen 1995]). Reduced ...