An Agent Based Method to Predict the Subsequent Concept in Concept Shifting Data Streams
Hossein Morshedlou, Ahamad Abdollahzadeh Barforoush · 2011
The advent of new application areas leads to intensive research on data streams. Due to concept shifts in the underlying data, maintaining a timely updated model has become one of the most challenging tasks. The data streams originate from the events of the real world so we think the associations among these events should exist among the concepts of the originated data stream too. Extraction of hidden associations among the concepts can be useful for prediction of subsequent concept in data stream. In this paper we present an agent-based method for classification of data streams with concept shifts. The agent creates a history from the concept changes and uses this history to have an intelligent behavior. The results of conducted experiments showed that the agent is proper for classification of concept shifting data streams.