A Wavelet-Genetic Programming Model for Predicting Short-Term and Long-Term Air Temperatures

A new conjunction wavelet-gene expression programming (WGEP) method for predicting air temperature values is proposed in this paper. The conjunction method combines the discrete wavelet and genetic programming methods. The daily and monthly air temperature data from two weather stations of Mahabad and Urmieh in Iran were used as case studies and the accuracy of the single gene expression programming (GEP) and wavelet-gene expression programming (WGEP) models were compared with each other. First, the daily air temperatures were used as inputs to the GEP and WGEP models to forecast one-, two- and three day as well as thirty-day ahead air temperatures. Then, the monthly air temperatures were used as inputs to the GEP and WGEP models to forecast one-month ahead air temperatures. The comparison results indicated that the WGEP model significantly increased the accuracy of single GEP model especially in forecasting long-term (thirty-day and one-month ahead) air temperatures. The thirty-day and one-month ahead air temperatures of the Mahabad Station were also estimated using the data of nearby Urmieh Station. It was found that the WGEP model performed much better than the single GEP model in cross-station application