Incremental Algorithm Driven by Error Margins

Incremental learning is an approach to deal with the classification task when datasets are too large or when new examples can arrive at any time. One possible approach uses concentration bounds (like Chernoff or Hoeffding bounds) to ensure that expansions are done when the number of examples supports the change. Two algorithms that use this approach are VFDT or IADEM. In this paper we extend the IADEM system in two directions: adding the ability to deal with continuous data and including the use of more powerful classification techniques at tree leaves. The proposed system, IADEMc, can incorporate and classify new information as the basic algorithms do, using shorter time per example. Another relevant property of our system is the ability to obtain a performance similar to a standard decision tree algorithm independently of the datasets size.

• Type: Conference
• Year: 2006
• Where: DIS
• Authors: Gonzalo Ramos-Jiménez, José del Campo-Ávila, Rafael Morales Bueno
• Keywords: Approach Uses Concentration, Classification Task, DIS 2006, Powerful Classification Techniques, Theoretical Computer Science