ABSTRACT
Weather forecasting is useful in several ways as in the view of agriculture, air traffic, ship routing, and various military applications. Weather prediction is proved to be a challenging task due to the frequent changes in atmospheric conditions. Therefore, a major requirement that arises in the research is to overcome the challenges imposed due to the chaotic nature of the weather. At present, the use of artificial intelligence and neural networks (NN) has been proven to be successful in many applications and tends to change the lives of human beings in many areas. For weather prediction, various models have been developed and a significant amount of research has been done by the researchers. However, considerably very limited research has been carried out in the meteorology area, and forecasting still relies on simulations via large and extensive computing resources. In this chapter, a comparison has been done between the autoregressive integrated moving average model and deep learning model convolutional recurrent neural network proportionally composed of convolutional neural network and recurrent neural network to forecast temperature. Extraction of spatial features has been done with the help of a deep learning model, which consists of a one-dimensional convolutional layer, and temporal features are extracted with the help of long short-term memory layers. Various attributes like temperature, humidity, wind speed, wind bearing, visibility, and air pressure are used in this work. Both the models are applied to hourly based temperature data set taken from Szeged, Hungary for the period 2006 to 2016. The performance of both the algorithms is computed using various performance metrics, like mean squared error, mean absolute error, R-squared error, and root-mean-square error.
