Building a Recommendation Engine: The XELOPES Library - Realtime Data Mining - page 251

MiningSettings

Settings

MiningAlgorithm
Specification

MiningInputStream MiningAlgorithm MiningModel

Input Output

MiningListener
MiningEvent
EventListenerList

Callback

Fig. 12.11 Main interfaces of MiningAlgorithm

The algorithm level represents special algorithm types of the functions. Many
algorithm types are predefined in CWM, and many have been added for XELOPES.
An example is the decisionTree algorithm which belongs to the function Classifi-
cation and represents a decision tree.

12.1.3.2 Algorithms

The abstract class MiningAlgorithm represents the data mining algorithm that
constructs a MiningModel .
Thus, MiningAlgorithm takes a mining input stream of the training data as input
and returns the mining model of the mining function as output. The training
parameters are passed through the mining settings on model-type level and mining
algorithm specification on algorithm level. Through a callback mechanism, the
training process can be monitored and controlled. The complete dataflow is shown
in Fig. 12.11 .
The central method of MiningAlgorithm is buildModel which runs the mining
algorithm and returns the mining model created by the algorithm. Internally,
buildModel calls the protected runAlgorithm method of the actual training process.
The buildModelWithAutomation method generates a mining model using tech-
niques for automatic parameter tuning, allowing to build mining models fully
automatically.
MiningAlgorithms owns a verify method which checks all parameters of the
algorithm class for correctness and completeness.