Online Sequential Prediction via Incremental Parsing: The Active LeZi Algorithm

Prediction is an important component in a variety of domains. Intelligent systems that can predict future events are better enabled to make more informed, and therefore more reliable, decisions. We investigate the use of compression methods for prediction with Active LeZi (ALZ), our online sequential prediction algorithm that can reason about the future in stochastic domains, with no domain-specific knowledge. The algorithm uses an information-theoretic approach and is experimentally analyzed using synthetic data, UNIX command data, and sequential data obtained from the MavHome smart home environment.

• Type: Journal
• Year: 2007
• Where: EXPERT
• Authors: Karthik Gopalratnam, Diane J. Cook
• Keywords: EXPERT 2007, Important Component, Prediction, Sequential Prediction Algorithm