Page 86 - DCAP603_DATAWARE_HOUSING_AND_DATAMINING

Data Warehousing and Data Mining

notes The artifcial neural networks which we describe are all variations on the parallel distributed
processing (PDP) idea. The architecture of each neural network is based on very similar building
blocks which perform the processing. In this unit we first discuss these processing units and
discuss diferent neural network topologies. Learning strategies as a basis for an adaptive
system

4.9 rule-based algorithms

One of the most well-studied methods for producing sets of classification rules from examples is
rule algorithms. They attempt to cover all instances of each class while excluding instances not
in the class. The main point is that covering algorithms (rule-based) work on a specific class at
a time, ignoring the rest of the classes. For instance, if a rule is desired to classify the weather as
warm, then the covering algorithm attempts to x in the statement.
If x, then class = warm,

With the condition that produces the best probability for the weather to be warm. Covering
algorithms follows these three steps:
1. Generate rule R on training data S
2. Remove the training data covered by rule R

3. Repeat the process
This method can be visualized in the 2D space of instances illustrated in Figure 4.7. First, a rule
is constructed to cover a’s by splitting the space vertically at x = 1.2 and then further splitting it
horizontally at y = 2.6, leading to the rule.
figure 4.7: covering algorithm Demonstration

If x > 1.2 AND y > 2.6, then class = a
Second, the following procedure is used to construct rules to cover b’s:
If x ≤ 1.2, then class = b
If x > 1.2 and y ≤ 2.6, then class = b
Note that one a is incorrectly covered by these rules, and more tests can be added to exclude that
a from b’s cover and include it in the a’s cover.

80 LoveLy professionaL university