ADAPTER: Adaptive Learning and Information Fusion for Online Classification Based on Evolving Big Data Streams

The aim of the project is to develop a new methodology for adaptive, distributed learning and information fusion from evolving data streams, based on the MapReduce paradigm. For the Map function, we will investigate adaptive learning methods of updating fuzzy approximate rules to assimilate new events and/or concept changes, given nonstationary and imbalanced data streams. For the Reduce function, we will develop an instance-based learning mechanism to reach more accurate results in the final decision about classification.