Machine Learning Based Weather Prediction Model for Short Term Weather Prediction in Sri Lanka

Weather forecasting in tropical countries like Sri Lanka is tremendously challenging since high temperature and sudden atmospheric circulations due to earth’s rotation causes turbulence effects which are difficult to be solved by prevailing numerical weather prediction methods. The aim of the current research is to introduce a weather prediction model for Sri Lanka, based on machine learning technology for making short term forecasts for weather attributes such as temperature and precipitation. This paper discusses making predictions on temperature for a selected weather station in Sri Lanka by implementing a multivariate Long Short-Term Memory Network (LSTM) which is trained on past weather observational data and evaluating the model’s performance using standard evaluation techniques. The Root Mean Squared Error (RMSE), Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE) values observed for the results of the LSTM model was extremely low, illustrating the ability of using machine learning models for identifying complex nonlinear patterns of past observational weather data and making accurate weather forecasts.