Context-Aware PDM (Coll-Stream)

Abstract

This chapter details how PDM can be extended to deal with the problem of concept drift. In such scenario, the goal is to learn an anytime classification model that represents the underlying concept from a stream of labeled records. Such a model is used to predict the label of the incoming unlabeled records. However, it is common for the underlying concept of interest to change over time and sometimes the labeled data available in the device is not sufficient to guarantee the quality of the results [66]. Therefore, we describe a framework that exploits the knowledge available in other devices to collaboratively improve the accuracy of local predictions.