The application of antigenic search techniques to time series forecasting

Time series have been a major topic of interest and analysis for hundreds of years, with forecasting a central problem. A large body of analysis techniques has been developed, particularly from methods in statistics and signal processing. Evolutionary techniques have only recently have been applied to time series problems. To date, applications of artificial immune system (AIS) techniques have been in the area of anomaly detection. In this paper we apply AIS techniques to the forecasting problem. We characterize a class of search algorithms we call antigenic search and show their ability to give a good forecast of next elements in series generated from Mackey-Glass and Lorenz equations. Categories and Subject Descriptors I.2.8 [Artificial Intelligence]: Problem Solving, Control Methods, and Search—dynamic programming General Terms Algorithms, Experimentation Keywords artificial immune systems, time series, forecasting, antigenic search

• Type: Conference
• Year: 2005
• Where: GECCO
• Authors: Ian Nunn, Tony White
• Keywords: Antigenic Search, GECCO 2005, Time Series, Time Series Problems