Building a Recommendation Engine: The XELOPES Library - Realtime Data Mining - page 253

classname ¼ "com.prudsys.pdm.Models.Sequential.Algo-
rithms.Seq.SequentialCycle"
version ¼ "1.0" >
< AlgorithmParameter name ¼ "minimumItemSize"
type ¼ "int"
value ¼ "1"
method ¼ "setM_minItemSize"
description ¼ "Minimum size for large items" / >
< AlgorithmParameter name ¼ "maximumItemSize"
type ¼ "int"
value ¼ "-1"
method ¼ "setM_maxItemSize"
description ¼ "Maximum size for large items" / >
< /AlgorithmSpecification >

■

Algorithm Types

Similar to MiningModel and MiningSettings , each class representing a type of data
mining algorithms extends MiningAlgorithm . For example, the general class of
association rule algorithms is AssociationRulesAlgorithm which extends
MiningAlgorithm . Again, this is the algorithm class associated with AssociationRu-
lesMiningModel and AssociationRulesSettings mentioned in before.
Along with all mining models and their mining settings, XELOPES provides the
associated algorithm classes containing the basic implementations.

Example 12.9 We give an example of the whole data mining process for sparse
grid classification (Chap. 7 ) . First, we build the sparse grid model. The training data
is contained in a CSV file whose path is specified in TRAIN_FILE . The target
attribute is supposed to be the last one. We apply (0,1) normalization to all numeric
attributes before we build the model. The resulting sparse grid model is written to
the PMML file SparseGridsModel.xml.
The Java code is given below:

// Open data source and get metadata:
MiningInputStream

inputData ¼ new

MiningCsvStream

( TRAIN_FILE );
inputData.open();
MiningDataSpecification metaData ¼ inputData.getMetaData
();

// Get target attribute (last one):
MiningAttribute targetAttribute ¼