Intelligent Miner® distinguishes
between a balanced and an unbalanced hierarchy. A balanced
hierarchy contains relations that start and end on two consecutive
levels. An unbalanced hierarchy contains relations that pass across
a hierarchy level.
The following figure shows a taxonomy with an unbalanced hierarchy:
Figure 1. Taxonomy with
an unbalanced hierarchy Level 0 Level 1 Level 2
Orange Juice |
>> Beverages |
Beer | |
|
>> Food
|
Fish |
|
Cheese |
If your hierarchy is unbalanced, you must make sure that a relation
that crosses hierarchy levels is in the same subset as the other relations
that end or start at the crossed hierarchy levels.
Usually, a taxonomy is defined by a number of recursive and non-recursive
categories. You must do the following tasks:
- Breaking up the set of relations in your taxonomy into subsets
- Describing these subsets by category maps of suitable types
Consider the following issues when you break up the set of relations
into subsets:
- Creating category maps for the balanced hierarchy
- You can divide the taxonomy that is shown below into the following
subsets:
- One subset holds the relations between hierarchy level 0 and 1
- The other subset holds the relations between hierarchy level 1
and 2
In this example, you must not split the relations into two subsets
because the relations comply with the restrictions.Figure 2. A taxonomy with a balanced hierarchy Level 0 Level 1 Level 2
Orange Juice |
>> Beverages |
Beer | |
|
>> Food
|
Tomato | |
| |
Cabbage >> Vegetables |
|
Carrot |
- Creating category maps for the unbalanced hierarchy
- You cannot break up the taxonomy that is shown in Figure 1 into
subsets because the relation Fish => Food crosses
the hierarchy level 1. Therefore, it must reside with the following
relations in the same subset:
- Orange Juice => Beverage
- Beer => Beverages
- Beverages => Food
The same applies to the relation Cheese => Food.
Consequently, all the relations of that taxonomy must reside in a
single set.
The following examples show the best combination of category maps
in various situations:
- Your taxonomy is balanced. The item names and the category names
are unique.
Use a recursive category map that holds the complete
set of relations.
You can also define several non-recursive
category maps, one for each relation between two consecutive hierarchy
levels.
- Your taxonomy is unbalanced. The item names and the category names
are unique.
Use a recursive category map that holds the complete
set of relations.
- Your taxonomy is balanced. The item names and the category names
are ambiguous among the hierarchy levels.
You must use non-recursive
category maps.
- Your taxonomy is unbalanced. The item names and the category names
are unique in the unbalanced parts of the taxonomy and ambiguous in
the balanced parts of the taxonomy.
You must separate the unbalanced
parts of the taxonomy from the balanced parts. Describe the unbalanced
parts by recursive category maps and the balanced parts by non-recursive
category maps.
You can define name mappings for your taxonomy. This is helpful
if your taxonomy uses identical item names and category names on different
levels or if categories use codes, for example, the code 7000 might
be used for Beverage.
You must define a name mapping for each category map individually.
The name mapping acts on all categories of the category map except
for the items and categories at the lowest hierarchy level of the
category map.
You must not define a name mapping for the lowest level of a category
map because the name mapping of the preceding category map acts on
the lowest level. The first category map that contains all relations
starting from hierarchy level 0 does not have a preceding category
map. In this case, the name mapping for the lowest level is defined
by the name mapping of the item field.
You define taxonomies when defining the logical data specification.
Note: A taxonomy must be acyclic; that is, a category must be neither
a direct member nor an indirect member of itself.