Comparative Analysis of Weather Forecasting Using LSTM, BiLSTM, and CLSTM Networks

Abstract

Weather prediction plays vital role in the evaluation of various processes and phenomenon for the betterment of society. Weather forecasting depends on many parameters hence, it’s a multivariable prediction problem. Behavior of data involved in the prediction are in majority of time stochastics hence, conventional mathematical approaches are failed to predict accurate data. Classical computational methods fail in predicting chaotic data due to its limitation to have know-how about behavior of data. Artificial Intelligence and Data Processing techniques is appealing investigators due to its efficiency to predict data based on its learning capability, without any know-how about the behavior of data. In this paper, Artificial Intelligence based approach LSTM, BiLSTM, and CLSTM are used to predict time series weather data prediction. Data considered are Pressure, Humidity, Temperature, and Maximum wind speed as input data. Neural Network based approaches are used to predict Average wind speed based on above inputs. The performance of above Artificial based approaches is compared to predict time series Average wind speed. MATLAB is used as software tool to implement Artificial Intelligence approaches. It has been observed that CLSTM algorithm outperformed other two algorithms LSTM and BiLSTM algorithm.