Temperature is one of the parameters that need to be considered because it is related to daily activities. Besides, the temperature is influenced by other parameters such as humidity, rainfall, and wind speed in the surrounding area. Data was obtained from Meteorology, Climatology, and Geophysics (BMKG) Bandung, West Java, from 2000-2019. This paper builds a model that can predict daily temperatures over the next three days with five classes, namely "Cold", "Cool", "Normal", "Warm" and "Hot" using Recurrent Neural Networks (RNN) and Long Short-Term Memory (LSTM). Before being predicted, pre-processing is needed to improve data quality consisting of interpolation, feature extraction, normalization, and segmentation. We use two optimization models, SGD and Adam. The results of this study prove that using Adam produces the best testing 90.92% for training data and 80.36% for test data. The amount of data used and the sharing of data can also affect the accuracy of the results obtained.
Daily Temperature Prediction Using Recurrent Neural Networks and Long-Short Term Memory
1 Views
1 Downloads