Modern OLAP platforms are capable of creating databases terabytes in size and present a significant challenge to the analyst with the goal of knowledge discovery. Artificial neural networks represent an aspect of machine learning that offers promise in this area. A neural map can learn to identify patterns in data of high dimensionality and a specific type of neural map, a neural tree classifier, can provide a hierarchical classification of the patterns identified. The investigation begins with a comparison of ...
Modern OLAP platforms are capable of creating databases terabytes in size and present a significant challenge to the analyst with the goal of knowledge discovery. Artificial neural networks represent an aspect of machine learning that offers promise in this area. A neural map can learn to identify patterns in data of high dimensionality and a specific type of neural map, a neural tree classifier, can provide a hierarchical classification of the patterns identified. The investigation begins with a comparison of two neural tree classifiers and continues by illustrating how their application can allow the identification of multi-dimensional areas of analytical interest in an OLAP database. Finally, a novel OLAP exception "explain" technique is outlined, enabled through the use of a neural tree classifier in conjunction with discovery-driven exploration.
Overview