C4.5: programs for machine learning
C4.5: programs for machine learning
Tracking Drifting Concepts By Minimizing Disagreements
Machine Learning - Special issue on computational learning theory
Learning in the presence of concept drift and hidden contexts
Machine Learning
Tracking Context Changes through Meta-Learning
Machine Learning - Special issue on multistrategy learning
The impact of changing populations on classifier performance
KDD '99 Proceedings of the fifth ACM SIGKDD international conference on Knowledge discovery and data mining
Incremental Learning from Noisy Data
Machine Learning
Mining Temporal Features in Association Rules
PKDD '99 Proceedings of the Third European Conference on Principles of Data Mining and Knowledge Discovery
Mining Surprising Patterns Using Temporal Description Length
VLDB '98 Proceedings of the 24rd International Conference on Very Large Data Bases
Classification of Customer Call Data in the Presence of Concept Drift and Noise
Soft-Ware 2002 Proceedings of the First International Conference on Computing in an Imperfect World
Hi-index | 0.00 |
In many application areas where databases are mined for classification rules, the latter may be subject to concept drift, that is change over time. Mining without taking this into account can result in severe degradation of the acquired classifier's performance. This is especially the case when mining is conducted incrementally to maintain knowledge used by an on-line system. The TSAR methodology detects and copes with drift in such situations through the use of a time stamp attribute, applied to incoming batches of data, as an integral part of the mining process. Here we extend the use of TSAR by employing more refined time stamps: first to individual batches, then to individual examples within a batch. We develop two new decision tree based TSAR algorithms, CD4 and CD5 and compare these to our original TSAR algorithm CD3.