Weather Forecasting Using Deep Learning Algorithms

Abstract

Weather forecasting aims to predict atmospheric conditions at a particular time and place. Timely alert of weather events is made possible through weather forecasting. For instance, accurate weather predictions enable us to offer early warning of natural disasters that significantly destroy both lives and property, such as cyclones, tsunamis, cloud bursts, etc. The aim of weather scientists has always been to provide accurate weather forecasts in a timely manner. Formerly, pattern recognition was frequently used for weather forecasting and all of such predictions have been lacking performance as far as accurate and precise forecasting is concern. As the conventional weather prediction techniques face a number of difficulties, such as: incomplete knowledge of physical processes, huge volumes of observational data are difficult to analyze, a need for strong computer resources, etc. To tackle these difficulties, this paper proposes to present an automatic weather forecasting model for short-range forecasting based on numerical and time series data using deep learning algorithms. This paper compares and assesses the performance of models created with various transfer functions in order to investigate the applicability of time series algorithms such as LSTM, GRU, and Bi-LSTM to develop an efficient and trustworthy nonlinear forecasting model for automatic weather analysis.