An Increment Decision Tree Algorithm for Streamed Data

Incremental (online) learning algorithms are methods for on-demand classification process from continuous streams of data. The main purpose is to deal with the classification task when original dataset is too large to process or when new instances of data arrive at any time. Moreover updating an existing model is (in many cases) much less expensive than to build a new one. This article presents a novel INEVOT algorithm for incremental decision tree induction from data streams. Because INEVOT is based on Evolutionary Algorithm it is possible to optimize different objectives at the same time. The experimental results indicate that proposed algorithm is powerful and promising. Provided solution can be easily adapted to nonstationary data streams.