TY - GEN
T1 - Multi resolution genetic programming approach for stream flow forecasting
AU - Rathinasamy, Maheswaran
AU - Khosa, Rakesh
PY - 2011/12/1
Y1 - 2011/12/1
N2 - Genetic Programming (GP) is increasingly used as an alternative for Artificial Neural Networks (ANN) in many applications viz. forecasting, classification etc. However, GP models are limited in scope as their application is restricted to stationary systems. This study proposes use of Multi Resolution Genetic Programming (MRGP) based approach as an alternative modelling strategy to treat non-stationaries. The proposed approach is a synthesis of Wavelets based Multi-Resolution Decomposition and Genetic Programming. Wavelet transform is used to decompose the time series at different scales of resolution so that the underlying temporal structures of the original time series become more tractable. Further, Genetic Programming is then applied to capture the underlying process through evolutionary algorithms. In the case study investigated, the MRGP is applied for forecasting one month ahead stream flow in Fraser River, Canada, and its performance compared with the conventional, but scale insensitive, GP model. The results show the MRGP as a promising approach for flow forecasting.
AB - Genetic Programming (GP) is increasingly used as an alternative for Artificial Neural Networks (ANN) in many applications viz. forecasting, classification etc. However, GP models are limited in scope as their application is restricted to stationary systems. This study proposes use of Multi Resolution Genetic Programming (MRGP) based approach as an alternative modelling strategy to treat non-stationaries. The proposed approach is a synthesis of Wavelets based Multi-Resolution Decomposition and Genetic Programming. Wavelet transform is used to decompose the time series at different scales of resolution so that the underlying temporal structures of the original time series become more tractable. Further, Genetic Programming is then applied to capture the underlying process through evolutionary algorithms. In the case study investigated, the MRGP is applied for forecasting one month ahead stream flow in Fraser River, Canada, and its performance compared with the conventional, but scale insensitive, GP model. The results show the MRGP as a promising approach for flow forecasting.
KW - Genetic Programming
KW - Multiscale Forecasting
KW - Stream flow
KW - Wavelet Analysis
UR - http://www.scopus.com/inward/record.url?scp=84555196626&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84555196626&partnerID=8YFLogxK
U2 - 10.1007/978-3-642-27172-4_84
DO - 10.1007/978-3-642-27172-4_84
M3 - Conference contribution
AN - SCOPUS:84555196626
SN - 9783642271717
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 714
EP - 722
BT - Swarm, Evolutionary, and Memetic Computing - Second International Conference, SEMCCO 2011, Proceedings
T2 - 2nd International Conference on Swarm, Evolutionary, and Memetic Computing, SEMCCO 2011
Y2 - 19 December 2011 through 21 December 2011
ER -