Automated Machine Learning Berbasis Pengetahuan Fisika untuk Model Prediksi Suhu dalam Kondisi Data Terbatas

Ade Hendini

doi:10.58794/jekin.v5i2.1578

Authors

Ade Hendini Universitas Bina Sarana Informatika

DOI:

https://doi.org/10.58794/jekin.v5i2.1578

Keywords:

AutoML, Physics-Informed Machine Learning, Prediksi Suhu, Data Terbatas, FLAML, H2O AutoML, NASA POWER

Abstract

Prediksi suhu merupakan bagian penting dalam berbagai sektor terutama pada sektor pertanian. Namun, keterbatasan data dan kompleksitas antar variabel menjadi tantangan utama pada pengembangan model prediktif suhu. Untuk mengatasi masalah tersebut, penelitian ini mengusulkan pendekatan baru dengan menggabungkan Automated Machine Learning (AutoML) dan pengetahuan fisika dalam satu kerangka kerja yang efisien dan domain-aware. Dengan membandingkan dua platform AutoML yaitu FLAML dan H2O AutoML. AutoML berperan mengotomatisasi pemilihan model, seleksi fitur, dan penyetelan parameter secara adaptif, sementara komponen berbasis pengetahuan fisika digunakan untuk membatasi ruang agar tetap konsisten dengan prinsip ilmiah. Dalam studi ini, data meteorologi harian seperti suhu, kelembapan, curah hujan, dan radiasi matahari diperoleh dari layanan NASA POWER. Model dikembangkan menggunakan pustaka FLAML dan H2O AutoML yang kemudian dievaluasi menggunakan metrik RMSE dan R². Hasil menunjukkan bahwa H2O AutoML memberikan performa prediksi yang lebih baik dengan nilai RMSE sebesar 0.6329 dan R² sebesar 0.6582, dibanding FLAML yang menghasilkan RMSE sebesar 0.7001 dan R² sebesar 0.4762. Visualisasi menunjukkan bahwa prediksi H2O lebih dekat terhadap nilai aktual dan mengikuti tren suhu dengan lebih akurat.

Downloads

Download data is not yet available.

References

R. K. Goel, C. S. Yadav, S. Vishnoi, and R. Rastogi, “Smart agriculture – Urgent need of the day in developing countries,” Sustainable Computing: Informatics and Systems, vol. 30, p. 100512, 2021, doi: https://doi.org/10.1016/j.suscom.2021.100512.

X. He, K. Zhao, and X. Chu, “AutoML: A survey of the state-of-the-art,” Knowl Based Syst, vol. 212, p. 106622, 2021, doi: https://doi.org/10.1016/j.knosys.2020.106622.

F. Hutter, L. Kotthoff, and J. Vanschoren, “The Springer Series on Challenges in Machine Learning Automated Machine Learning Methods, Systems, Challenges.” [Online]. Available: http://www.springer.com/series/15602

V. K. Quy et al., “IoT-enabled smart agriculture: architecture, applications, and challenges,” Applied Sciences, vol. 12, no. 7, p. 3396, 2022.

L. G, R. C, and G. P, “An automated low cost IoT based Fertilizer Intimation System for smart agriculture,” Sustainable Computing: Informatics and Systems, vol. 28, p. 100300, 2020, doi: https://doi.org/10.1016/j.suscom.2019.01.002.

M.-A. Zöller and M. F. Huber, “Benchmark and Survey of Automated Machine Learning Frameworks,” J. Artif. Int. Res., vol. 70, pp. 409–472, May 2021, doi: 10.1613/jair.1.11854.

S. A. Mezzasalma, “On the Extension of a Physical Body in Classical Motion. An Analogy for a Pseudo-Velocity Concept and Wiener’s Process in (Ideal) Polymer Solutions,” 2020. [Online]. Available: https://arxiv.org/abs/2009.06105

G. E. Karniadakis, I. G. Kevrekidis, L. Lu, P. Perdikaris, S. Wang, and L. Yang, “Physics-informed machine learning,” Nature Reviews Physics, vol. 3, no. 6, pp. 422–440, 2021, doi: 10.1038/s42254-021-00314-5.

M. Raissi, P. Perdikaris, and G. E. Karniadakis, “Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,” J Comput Phys, vol. 378, pp. 686–707, 2019, doi: https://doi.org/10.1016/j.jcp.2018.10.045.

S. Liao, T. Xue, J. Jeong, S. Webster, K. Ehmann, and J. Cao, “Hybrid thermal modeling of additive manufacturing processes using physics-informed neural networks for temperature prediction and parameter identification,” Comput Mech, vol. 72, no. 3, pp. 499–512, Jan. 2023, doi: 10.1007/s00466-022-02257-9.

T. X. Nghiem et al., “Physics-Informed Machine Learning for Modeling and Control of Dynamical Systems,” 2023. [Online]. Available: https://arxiv.org/abs/2306.13867

X. Wu et al., “AutoPINN: When AutoML Meets Physics-Informed Neural Networks,” 2022. [Online]. Available: https://arxiv.org/abs/2212.04058

G. Guo, H. You, C. Li, Z. Tang, and O. Li, “A Physics-Informed Automatic Neural Network Generation Framework for Emerging Device Modeling,” Micromachines (Basel), vol. 14, no. 6, 2023, doi: 10.3390/mi14061150.

C. Wang, Q. Wu, X. Liu, and L. Quintanilla, “Automated Machine Learning & Tuning with FLAML,” in Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, in KDD ’22. New York, NY, USA: Association for Computing Machinery, 2022, pp. 4828–4829. doi: 10.1145/3534678.3542636.