Time series forecasting of air temperature using an intelligent hybrid model of genetic algorithm and neural network

Owing to climate change, global warming, and contemporary droughts, temperature forecasting, as one of the most influential climatic parameters, produces a well-suited opportunity for executives to plan and provide the necessary preparations. The matter of time series forecasting of air temperature is one of the most intriguing issues in climate investigations. In this article, an intelligent hybrid model is presented to predict the time series of air temperature. This paper uses the idea of practicing the feature selection model based on the genetic algorithm (GA) to determine the input variables of the model and the high forecasting power of the neural network. The recommended model used the structure of the Autoregressive time series model. But, the problem of selecting the delay of the time series when they should be used in the model was done using genetic algorithm. Finally, the selected delays were used as input of the neural network model. The average monthly air temperature of Tabriz and Kermanshah stations throughout the statistical period 1980-2010 was used to assess the proposed model. The performance of the suggested model was compared with neural network models that do not use the feature selection method. The results corroborated the high accuracy of the developed model compared to the other models, indicating the significance of the problem of feature selection in predicting time series.