Prediksi Cuaca di Indonesia Menggunakan Optimized Lag-Llama Berdasarkan Data BMKG Kemayoran dan Kualanamu

Syamsuri, Kgs M Nur (2026) Prediksi Cuaca di Indonesia Menggunakan Optimized Lag-Llama Berdasarkan Data BMKG Kemayoran dan Kualanamu. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6025232003-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Akurasi dan kelengkapan informasi dalam prediksi cuaca memainkan peran krusial dalam mendukung pengambilan keputusan berbasis risiko, khususnya dalam sistem peringatan dini cuaca ekstrem serta pengelolaan sumber daya alam. Kompleksitas dan ketidakpastian sistem atmosfer menuntut pendekatan prediktif yang mampu merepresentasikan ketidakpastian tersebut secara eksplisit. Studi ini mengkaji potensi model Lag-Llama yang sebuah model fondasi berbasis arsitektur transformer, dalam menghasilkan prediksi probabilistik terhadap data cuaca di Indonesia. Model ini dirancang untuk memproyeksikan variabel-variabel meteorologis utama seperti suhu, kelembapan, dan tekanan udara dalam rentang waktu 24 jam ke depan. Untuk meningkatkan performa model, dilakukan optimasi hyperparameter menggunakan Optuna, dengan fokus pada parameter utama seperti panjang konteks (context length) dan lag. Evaluasi performa dilakukan dengan menggunakan metrik MAE, MAPE, dan Continuous Ranked Probability Score (CRPS) guna menilai akurasi serta kualitas estimasi distribusi probabilistik yang dihasilkan. Temuan dari penelitian ini diharapkan memberikan kontribusi signifikan dalam pengembangan sistem prediksi cuaca yang lebih presisi dan adaptif di Indonesia, sekaligus mendukung strategi mitigasi perubahan iklim melalui penerapan teknologi kecerdasan buatan (AI).
=======================================================================================================================================
Accurate and informative weather forecasting is vital for risk-informed decision-making, particularly in early warning systems for extreme weather and the management of natural resources. However, the inherent complexity and uncertainty of atmospheric dynamics necessitate predictive models that can explicitly account for such uncertainty. This study investigates the potential of the Lag-Llama model, a foundation model based on the transformer architecture for generating probabilistic weather forecasts in Indonesia. The model is adapted to predict key meteorological variables, including temperature, humidity, and air pressure, over forecast horizons of 24 hours. To enhance model performance, hyperparameter optimization is conducted using Optuna, with a focus on critical parameters such as context length and lag. Model evaluation employs metrics including Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and the Continuous Ranked Probability Score (CRPS) to assess both predictive accuracy and the quality of the probabilistic distribution outputs. The outcomes of this research are expected to contribute meaningfully to the development of more adaptive and precise weather forecasting systems in Indonesia, while also supporting climate change mitigation strategies through the application of artificial intelligence technologies.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	cuaca indonesia, CRPS, hyperparameter tuning, lag-llama, optuna, prediksi probabilistik, transformer, indonesian weather, probabilistic forecasting
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence T Technology > T Technology (General) > T174 Technological forecasting T Technology > TD Environmental technology. Sanitary engineering > TD171.75 Climate change mitigation
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User:	Kgs M Nur Syamsuri
Date Deposited:	26 Jan 2026 05:53
Last Modified:	26 Jan 2026 05:53
URI:	http://repository.its.ac.id/id/eprint/130315

Actions (login required)

View Item