Performance Analysis Of Autoregressive Integrated Moving Average Model And Deep Learning Long Short-Term Memory Model For Weather Data Forecasting

Abstract

Weather is an aspect that cannot be separated from all activities carried out by humans, so information about the weather is very important. To meet the need for this information, it is necessary to do forecasting. Each data has its own characteristics, and choosing the right forecasting method is very important. The Autoregressive Integrated Moving Average (ARIMA) method is one of the popular statistical methods used in forecasting time-series data. Long Short-Term Memory (LSTM) is a modern deep learning algorithm model that is most suitable for forecasting time-series data. In this study, an analysis was carried out to compare the traditional ARIMA method and the deep learning model, namely LSTM, in forecasting weather data in Manado city to see the best forecasting model that can be used. The results of this study indicate that in terms of the accuracy of the 18 tests performed, the LSTM forecasting model is superior to the ARIMA model as measured by Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE). In terms of computational time in making forecasting models for 6 weather data attributes, the LSTM model is faster than the ARIMA model.