Automating Data Preprocessing with DMPML and KDDML

Abstract

This paper presents a graphical application for the Data Mining Preparation Markup Language (DMPML), which is an XML application designed to represent the data preparation phase of the KDD process. DMPML supports the reuse of data preprocessing directives using XSLT to map raw data into data ready to be used by many data mining algorithms. The application presented here, DMPML-TS, automates the data preparation phase, speeding up the codification and transformation of data, and providing support to facilitate the use of different data mining algorithms in the same and/or similar data, based on their codification stored in separate XML documents. This paper also presents improvements made to DMPML like the adoption of XRFF for input and output data and the use of only one XSLT file for data transformation. We also present the integration of DMPML-TS and KDDML, an XML language used to represent data, mining models, and queries.