Time Series Weather Data Forecasting Using Deep Learning
Pau Suan Mung, Sabai Phyu
Abstract
Weather forecasting is an interesting research in a number of applications and has great attention of researchers from various research communities due to its effect on the daily life of human globally. Through the years, researchers used sliding window and different machine learning techniques for this purpose. Weather data is in the form of time series data. Time series data can be forecasted using regression. For better accuracy and performance, deep learning becomes popular for prediction. The aim of this research work is to predict weather attributes such as minimum temperature, maximum temperature, humidity and wind speed using different deep learning methods, namely convolutional neural networks (CNN), long short-term memory (LSTM) and ensemble of CNN and LSTM (CNN-LSTM). Dataset including daily weather conditions from past two decades of Yangon Region, Myanmar is used as case study for this research and one-month weather conditions will be predicted. Root mean square error (RMSE) is used for comparison of the performance of these methods. The experiment shows that ensemble CNN-LSTM model has better performance than other individual methods.