One is rule-based classification and one is biased classification or naive biased classification.

Let's first talk about rule-based classification.

I just think there are no questions. If there are questions, please interrupt me if necessary.

Rule-based classification is pretty close to decision tree classification because you can derive rules from decision trees.

A rule in the sense of classification is represented by any if-then rule.

So, for example, we might have if age below 30, student and student yes, when bias computer.

That's something you should already know from back here.

This tree. If age below 30, below equal 30 and student yes, then this is also yes.

So basically, we can derive if-then rules from a decision tree.

The big advantage of if-then rules is that they are easier to read because you can easier just read a line and understand that line.

We, of course, have multiple possibilities to work with a rule.

If we have a rule set, we might trigger more than one rule and there might be a rule if age lower than 30 and a second rule if age lower than 45.

Then, of course, both rules would trigger for a person who is 25 because both are true.

And also, there might be the possibility that no rules, no rule triggers.

Again, if we have these two rules and we are a person above the age of 45, for example, 50, then we might also trigger no rule.

These two things might happen and depending on which problem occurs, we might have to resolve that conflict.

This can be done, especially for triggering more than one rule, that we first order these rules according to a priority list.

Most often that is done by size ordering.

So the strictest rule is applied first.

And then, if this rule is satisfied, we don't have to check further rules.

We can also order our rules glass-based.

So, for example, all rules regarding age are ordered together.

So we see if a rule applies or a table might apply to two rules or more.

And we also might have a custom priority scaling.

If we have a custom priority scaling, of course, we have to apply all rules according to that priority ordering.

However, that's also true for all the other orders.

If we want to avoid triggering multiple rules, we, of course, have to stick to our ordering.

Whether it's derived from strictness or it's derived from classes or it's derived from any other rule we made up.

The second case, the second problem was if there's no rule that is triggered.

And that's pretty simple to solve because we just can set up a default rule.

If you think back to decision trees, a default rule might be the most often occurring class.

That might be a default rule.

So, for example, if we have two classes, yes and no, and yes is occurring 70 percent of cases overall and 30 percent of cases are no,

there would be a case to be made to set yes as a default rule to become part of that.

This default rule should then, of course, be related as the last rule.

And if only if no other rules apply, then this rule is triggered.

But I guess that's pretty obvious.

We also already talked about rules can be extracted from a decision tree.

In our decision tree example from the first section, we had an age student graduating decision tree.

And for example, we can extract age below 30 and student no equals vice computer no from that tree.

So this path down here can be rewritten as an event rule, of course.

Question, can we extract age below 30 equals no from this decision tree?

Who is for yes? Raise your hands.

So who thinks this rule can be extracted from the tree?

Age below 30 leads to no.

Who's thinking that this can be extracted?

Raise your hand. Who's against it can be extracted?

No risen hands at all.

So no opinion.

Maybe I should write down the rule to avoid my accent breaking that question.