Building a Recommendation Engine: The XELOPES Library - Realtime Data Mining - page 239

Warehouse Process Warehouse Operation

Management

Transformation OLAP Data

Mining

Business
Nomenclature

Information
Visualization

Analysis

Relational Record Multidimensional XML

Resource

Business
Information

Data
Types

Software
Deployment

Key
Indexes

Expres
sions

Type
Mapping

Foundation

Core Behavioral Relationships Instance

Object

Fig. 12.5 CWM packages used in MiningDataSpecification

12.1.2 The Building Blocks of the Library

12.1.2.1 The Basis: MiningDataSpecification and MiningAttributes

The class MiningDataSpecification represents the basis of a space. Thus, this is the
most important class of XELOPES. Often, this class is simply referred to as
metadata of the mining what is the equivalent of calling it the basis.
The basis vectors of MiningDataSpecification are the MiningAttributes representing
the attributes. Therefore, MiningDataSpecification extends the CWM class Class ,and
MiningAttribute extends the CWM class Attribute (Fig. 12.3 ).
There are two basic types of mining attributes extending the abstract class
MiningAttribute : NumericAttribute for numeric attributes like age, income, and
time and CategoricalAttribute for categorical attributes like names, IDs, and types.
The elements of a numeric attribute are real numbers. The elements of a categorical
attribute are the categories which are represented by the Category class.

Example 12.1
// Create category 'knife':
Category catKnife ¼ new Category("knife"); ■

The categories of a categorical attribute are stored in an array of Categorical
Attribute . Unlike as for the straightforward NumericAttribute , the mathematical
nature of the CategoricalAttribute is rather ambivalent: it can be interpreted as
one or a set of multiple numeric attributes. In the last case (e.g., binning),
CategoricalAttribute represents a basis itself, with the Category -s as basis
vectors. Thus, the set of categories is also called the basis or the metadata of
the categorical attribute. Each category of a categorical attribute can be mapped
to a unique real number (usually an integer) which is called the key of this
category. This establishes a mutually unique mapping between the categories and