Five interesting rules from Foil and Census Income dataset.
Full Set of Rules (if hardcopy, please see attached copy) formatted to look pretty.

Interesting statistics for attributes

      So, it is easy to see that Foil only used continuous attributes to choose the rules. could this have anything to do with the theory constants, or lack thereof?

Some interesting rules and points:

1. education-num>12, capital-gain>6849, age<=62
   • One has to wonder about the age restriction for this rule.


2. age>33, hours-per-week>39, education-num>14, capital-gain>4064
   • This rule shows off one of the great advantages of rule based systems. It's logic is very easy to follow.


3. age>33, capital-gain>5013
   • This rule is very general.


4. education-num>12, capital-gain<=3103, capital-gain>2977, hours-per-week<=47
   • This is similar to rule 1, but instead of age it uses a range for capital gain and hour per week worked.


5. age>29, education-num>9, hours-per-week>44, capital-loss>1741, age<=31, hours-per-week<=55
   • With capital gain getting so much press, it was only fair to include at least one rule that used capital loss. The very small age range and slack education requirements in this rule are odd.

Notice how similar rules 2 and 3 are.