A multistrategy approach to classification learning in databases

摘要

This paper proposes a hybrid classification learning system for databases that integrates rule induction and lazy learning. For rule induction learning, we use an entropy function based on Hellinger divergence to measure the amount of information each inductive rule contains. For lazy learning, we also use the Hellinger measure to automatically generate attribute weights and to compute similarities between data values of non-numeric data types. Our system has been implemented and tested extensively on a number of well-known machine learning data sets. The performance of our system was favorable compared to those of other well-known classification learning techniques based on monostrategic learning methods.