HYPERPARAMETER TUNING LSTM SEBAGAI ESTIMATOR SENSOR RELATIVE HUMIDITY PADA AUTOMATIC WEATHER STATION BERBASIS SIMULATED ANNEALING

HYPERPARAMETER TUNING LSTM AS RELATIVE HUMIDITY SENSOR ESTIMATOR ON AUTOMATIC WEATHER STATION BASED ON SIMULATED ANNEALING

Authors

  • Naufal Ananda BMKG
  • Haryas Subyantara Wicaksana
  • Yusuf Giri Wijaya
  • Rhakamerta Hijazi

Keywords:

Hyperparameter tuning, Simulated Annealing, Relative Humidity

Abstract

Pengukuran kelembapan udara relatif (RH) sebagai salah satu besaran cuaca dilakukan di lapisan permukaan menggunakan Automatic Weather Station (AWS). Pada tahun 2020, sensor RH AWS  masih memiliki 7% tingkat unavailability karena kerusakan pencatu daya, kerusakan sensor serta gangguan jaringan komunikasi. Pada penelitian ini, dirancang estimasi nilai sensor RH AWS yang dapat dijadikan alternatif terhadap unavailability data RH AWS. Optimasi performa algoritma LSTM sebagai estimator RH dapat dilakukan melalui hyperparameter tuning berbasis simulated annealing (SA). Data diambil dari output sensor RH AWS Pemalang, Jawa Tengah. Model LSTM terbaru (LSTM-SA) selanjutnya digunakan sebagai estimator data sensor RH AWS Pemalang. Hasil estimasi sensor RH AWS Pemalang menggunakan model LSTM-SA kemudian dikomparasi terhadap model LSTM tanpa SA. Jumlah neuron optimal berdasarkan algoritma simulated annealing yaitu 48 neuron per hidden layer. Batch size optimal berdasarkan algoritma simulated annealing yaitu 21. Nilai RMSE 0,015% lebih rendah dibanding nilai RMSE awal tanpa hyperparameter tuning terhadap batch size. Algoritma LSTM-SA mampu mengoptimasi hyperparameter algoritma LSTM dengan menurunnya nilai RMSE. Nilai error yang dihasilkan masih kurang dari 3 %RH sesuai ketentuan dokumen World Meteorological Organization (WMO) No.8.

 

References

BMKG (2014) : Peraturan Kepala Badan Meteorologi Klimatologi dan Geofisika Nomor 7 Tahun 2014 tentang Standar Teknis dan Operasional Pemeliharaan Peralatan Pengamatan MKG.

World Meteorological Organization (2018) : WMO No.8 Guide to Instruments and Methods of Observation.

Dr Bushra Shamshad, M Zubair Khan, Zara Omar, “Modeling and Forecasting Weather Parameters using ANN-MLP, ARIMA and ETS model: A case study for Lahore, Pakistan”, International Journal of Scientific & Engineering Research,Volume 10, Issue 4, 2019.

Mohammad Shad, Y. D. Sharma Abhishek Singh, “Forecasting og Monthly Relative Humidity in Delhi, India, using SARIMA and ANN models”, Modeling Earth Systems and Environment, Springer,2022.

Antonios Parasyris, George Alexandrakis, Georgios V. Kozyrakis, Katerina Spanoudaki dan Nikolaos A. Kampanis, “Predicting Meteorological Variables on Local Level with SARIMA, LSTM and Hybrid Techniques”, Atmosphere, 2022, https://doi.org/10.3390/atmos13060878.

Chun-Wei Tsai, Chien-Hui Hsia, Shuang-Jie Yang, Shih-Jui Liu, Zhi-Yan Fang, “Optimizing hyperparameters of deep learning in predicting bus passengers based on simulated annealing”, Applied Soft Computing Journal 88, 2020.

Sergio Ledesma, Miguel Torres, Donato Hernández, Gabriel Aviña, and Guadalupe García, “Temperature cycling on simulated annealing for neural network learning”, Advances in Artificial Intelligence, hal. 161–171, 2007.

Jussi Kalliola, Jurgita Kapočiūtė-Dzikienė, dan Robertas Damaševičius, “Neural network hyperparameter optimization for prediction of real estate prices in Helsinki”, PeerJ Comput. Sci. 7, 2021.

Nikolaos Gorgolis, Ioannis Hatzilygeroudis, Zoltan Istenes, Lazlo – Grad Gyenne, “Hyperparameter Optimization of LSTM Network Models through Genetic Algorithm”, Conference Paper, 2019.

Bahareh Nakisaa, Mohammad Naim Rastgoo, Andry Rakotonirainy, Frederic Mairea, Vinod Chandran, “Long Short Term Memory Hyperparameter Optimization for a Neural Network Based Emotion Recognition Framework”, IEEE Access, 2022.

23

Downloads

Published

2023-02-02

How to Cite

Ananda, N., Wicaksana, H. S., Yusuf Giri Wijaya, & Rhakamerta Hijazi. (2023). HYPERPARAMETER TUNING LSTM SEBAGAI ESTIMATOR SENSOR RELATIVE HUMIDITY PADA AUTOMATIC WEATHER STATION BERBASIS SIMULATED ANNEALING: HYPERPARAMETER TUNING LSTM AS RELATIVE HUMIDITY SENSOR ESTIMATOR ON AUTOMATIC WEATHER STATION BASED ON SIMULATED ANNEALING. Buletin Meteorologi, Klimatologi Dan Geofisika, 3(1), 35–43. Retrieved from http://balai2bmkg.id/index.php/buletin_mkg/article/view/66