Time Series Temperature Forecasting by using ConvLSTM Approach,
Case Study in Jakarta
Dublin Core
Title
Time Series Temperature Forecasting by using ConvLSTM Approach,
Case Study in Jakarta
Case Study in Jakarta
Subject
Temperature Forecasting, Machine Learning, MLP, LSTM, ConvLSTM
Description
Climate change has occurred in several countries, especially in tropical countries such as Indonesia. It causes extreme
temperature changes in several Indonesian areas, especially Jakarta, one of the world's most populated cities. The population
of Jakarta causes the activities carried out by residents to be disturbed by extreme temperature changes. In addition, drastic
temperature changes also affect the energy consumption used by residents. Therefore, it is necessary to predict temperature to
determine future temperature conditions so that residents can plan their activities. Temperature forecast can be done in several
ways, one of which uses a machine learning approach. This research uses a deep learning model called the Convolutional
Long Short-Term Memory (ConvLSTM). Moreover, we also compare the model with Multi-Layer Perceptron (MLP), and Long
Short-Term Memory (LSTM). We use temperature data taken from the ERA-5 period years 2018 to 2020 located in Kemayoran,
Jakarta, Indonesia. This research aims to investigate the accuracy of short-term temperature forecasting by using these three
models. The model is built to predict short-term temperatures for 1, 3, and 7 days ahead. The performance of the three methods
is measured by calculating the Root Mean Square Error (RMSE), Mean Square Error (MAE), and Coefficient Correlation
(CC). The result shows that the LSTM performs better than the other methods to forecast 1, 3, and 7 days, i.e., with the lowest
RMSE, MAE, and higher CC.
temperature changes in several Indonesian areas, especially Jakarta, one of the world's most populated cities. The population
of Jakarta causes the activities carried out by residents to be disturbed by extreme temperature changes. In addition, drastic
temperature changes also affect the energy consumption used by residents. Therefore, it is necessary to predict temperature to
determine future temperature conditions so that residents can plan their activities. Temperature forecast can be done in several
ways, one of which uses a machine learning approach. This research uses a deep learning model called the Convolutional
Long Short-Term Memory (ConvLSTM). Moreover, we also compare the model with Multi-Layer Perceptron (MLP), and Long
Short-Term Memory (LSTM). We use temperature data taken from the ERA-5 period years 2018 to 2020 located in Kemayoran,
Jakarta, Indonesia. This research aims to investigate the accuracy of short-term temperature forecasting by using these three
models. The model is built to predict short-term temperatures for 1, 3, and 7 days ahead. The performance of the three methods
is measured by calculating the Root Mean Square Error (RMSE), Mean Square Error (MAE), and Coefficient Correlation
(CC). The result shows that the LSTM performs better than the other methods to forecast 1, 3, and 7 days, i.e., with the lowest
RMSE, MAE, and higher CC.
Creator
Faishal Raihanur Rasyid1
, Didit Adytia2
, Didit Adytia2
Publisher
Telkom University
Date
22-08-2022
Contributor
Fajar bagus W
Format
PDF
Language
Indonesia
Type
Text
Files
Collection
Citation
Faishal Raihanur Rasyid1
, Didit Adytia2, “Time Series Temperature Forecasting by using ConvLSTM Approach,
Case Study in Jakarta,” Repository Horizon University Indonesia, accessed June 6, 2025, https://repository.horizon.ac.id/items/show/9208.
Case Study in Jakarta,” Repository Horizon University Indonesia, accessed June 6, 2025, https://repository.horizon.ac.id/items/show/9208.