Bitmap Index-Based Decision Trees

In this paper we propose an original approach to apply data mining algorithms, namely decision tree-based methods, taking into account not only the size of processed databases but also the processing time. The key idea consists in constructing a decision tree, within the DBMS, using bitmap indices. Indeed bitmap indices have many useful properties such as the count and bit-wise operations. We will show that these operations are efficient to build decision trees. In addition, by using bitmap indices, we don’t need to access raw data. This implies clear improvements in terms of processing time. Key words : Bitmap indices, databases, data mining, decision trees, performance.

• Type: Conference
• Year: 2005
• Where: ISMIS
• Authors: Cécile Favre, Fadila Bentayeb
• Keywords: Bitmap Indices, Data Mining, Decision Trees, ISMIS 2005