A Software Architecture for Data Mining Environment

摘要: buildClassifier (Contexte) classifyExemple ( Example ex) Instances Instance input Option value : String name: Description: has Fig. 7. A partial view of WEKA with the new classification algorithms In extension is done through inheritance. The choice this approach supposes that are written in Java. management heterogeneity reuse other programming languages relies on competence developer. So next step will consist to provide mechanism which could easy adaptation heterogeneous algorithms. 5.3.2 Managing interoperability platform implements into environment (classifier, cluster and association rule) use as parameter, data under an owner format (ARFF, Attribute Relation File Format). They produce at output textual unstructured descriptions. absence explicit description characteristics models thus built does not allow extracted knowledge. From flat structure be main each type model. case rules, model consists all rules found by algorithm. importance rule evaluated metric used general it about confidence. Each two sets items (called Itemset) characterized their support. first being antecedent or premise second conclusion consequence. One Item indicates article a pair attribute/value source. sequential motifs, rather made up set sequences Sequence). sequence chronological series Itemsets. Three major concepts permit characterize independently platforms algorithms: items, ItemSets rules. same reasoning can applied easily types order extract characteristics. Figura 9 presents process been export from algorithm associate family WEKA.