Pentaho + R: An Integral View for Multidimensional Prediction Models

The integration of multidimensional data and machine learning seems to be natural in the area of business intelligence. On-Line Analytical Processing (OLAP) tools are frequent in this area where the data are usually represented in multidimensional datamarts and data mining tools are integrated in some of these tools. However, the efforts for a full integration of data mining and OLAP tools have not been as common as originally expected. Nowadays, this integration is mostly carried out on source code, implementing solutions that perform (i) all the operations on multidimensional data as well as (ii) the data mining algorithms to extract knowledge from these data. Hence, there now exists an important distinction between implementation-based developments where the entire solution is implemented on source code and OLAP-tool-based developments where (at least) the operations on multidimensional data are performed using an OLAP tool. This work analyses these two alternatives in cost-effective terms, performing an experimental analysis on a multidimensional problem and discussing when each approach seems to excel the other.