Heterogeneous data reduction model for payment request file of direct debit processes

This paper presents a proposed model regarding Heterogeneous Data Reduction. The model reduces data over a heterogeneous environment through feature selection/extraction. The feature is selected/extracted directly from its data source and prepared without an initial integration for all data sources. After that the selected/extracted prepared feature is integrated into a new reduced data set Feature selection/extraction is made according to business requirements, domain expert feedbacks, and the organization´s Service Level Agreement and Corporate Household to give high accuracy results. The proposed model is built by hybrid data reduction techniques: Stepwise Backward Elimination, Stepwise Forward Selection and Decision Tree Induction. Such proposed model building depends on the CRoss Industry Standard Process model of data mining as a reference model The proposed model works with any kind of data types. The model applies to real telecommunication data relating to the Direct Debit processes. It is used to produce a Standard Converted Reduced Payment Request File to just keep on the important attributes. The model helps to cut down the user work time to generate that new data set.