Mukti, Muhammad Maulana (2026) Prediksi Tingkat Kekeringan Pertanian Menggunakan Model Bidirectional Long Short-Term Memory (BiLSTM) Berbasis Data NASA POWER dan U.S. Drought Monitor. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5026221201-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5026221201-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy |
Abstract
Kekeringan merupakan salah satu permasalahan utama pada sektor pertanian karena berdampak langsung terhadap produktivitas tanaman, ketersediaan air, serta stabilitas ketahanan pangan. Upaya mitigasi melalui sistem prediksi dini masih menghadapi kendala akibat keterbatasan data lokal yang lengkap dan berkesinambungan. Oleh karena itu, penelitian ini mengembangkan model prediksi tingkat kekeringan berbasis Bidirectional Long Short-Term Memory (BiLSTM) dengan memanfaatkan data agroklimatologi NASA POWER dan data klasifikasi bulanan U.S. Drought Monitor (USDM) pada wilayah Kansas dan Nebraska. Data diproses melalui tahapan pembersihan, rekayasa fitur, normalisasi, pembentukan time-series sequence, serta penanganan ketidakseimbangan kelas menggunakan kombinasi Random Over Sampling (ROS) dan Categorical Focal Loss. Selanjutnya dilakukan serangkaian eksperimen untuk mengevaluasi pengaruh berbagai komposisi fitur terhadap performa model BiLSTM. Evaluasi dilakukan menggunakan metrik Accuracy, Precision, Recall, F1-Score, dan Macro F1-Score. Hasil penelitian menunjukkan bahwa model BiLSTM mampu memprediksi tingkat kekeringan secara konsisten pada kedua wilayah penelitian. Selain itu, komposisi fitur terbukti memberikan pengaruh yang signifikan terhadap performa model, yaitu ketika fitur riwayat kekeringan memberikan kontribusi paling dominan, sedangkan proses seleksi fitur yang tepat mampu mengurangi jumlah fitur tanpa menurunkan performa secara berarti. Penelitian ini menunjukkan bahwa keberhasilan prediksi tingkat kekeringan tidak hanya dipengaruhi oleh arsitektur model, tetapi juga oleh kualitas dan relevansi fitur yang digunakan, sehingga dapat menjadi dasar dalam pengembangan sistem peringatan dini kekeringan berbasis data agroklimatologi.
======================================================================================================================================
Agricultural drought is one of the major challenges affecting crop productivity, water availability, and food security. Developing an effective early warning system remains challenging due to the limited availability of comprehensive and continuous local climate data. To address this issue, this study developed a drought level prediction model based on the Bidirectional Long Short-Term Memory (BiLSTM) architecture using agroclimatological data from NASA POWER and monthly drought classification data from the U.S. Drought Monitor (USDM) for the Kansas and Nebraska regions. The dataset was processed through data cleaning, feature engineering, normalization, time-series sequence generation, and class imbalance handling using a combination of Random Over Sampling (ROS) and Categorical Focal Loss. A series of experiments was conducted to evaluate the influence of different feature compositions on the performance of the BiLSTM model. Model performance was assessed using Accuracy, Precision, Recall, F1-Score, and Macro F1-Score. The results demonstrate that the proposed BiLSTM model consistently predicts drought levels across both study regions. Furthermore, feature composition was found to have a substantial impact on model performance, with drought history features providing the greatest contribution, while appropriate feature selection reduced the number of input features without significantly degrading predictive performance. These findings indicate that successful drought prediction depends not only on the deep learning architecture but also on the quality and relevance of the input features, providing a practical foundation for the development of agroclimatology-based drought early warning systems.
| Item Type: | Thesis (Other) |
|---|---|
| Uncontrolled Keywords: | Kekeringan Pertanian, Time Series, BiLSTM, NASA POWER, U.S. Drought Monitor, Agricultural Drought, Time Series, BiLSTM, NASA POWER, U.S. Drought Monitor. |
| Subjects: | T Technology > T Technology (General) > T174 Technological forecasting T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T57.62 Simulation T Technology > T Technology (General) > T57.84 Heuristic algorithms. |
| Divisions: | Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis |
| Depositing User: | Muhammad Maulana Mukti |
| Date Deposited: | 28 Jul 2026 04:22 |
| Last Modified: | 28 Jul 2026 04:22 |
| URI: | http://repository.its.ac.id/id/eprint/138534 |
Available Versions of this Item
- Prediksi Tingkat Kekeringan Pertanian Menggunakan Model Bidirectional Long Short-Term Memory (BiLSTM) Berbasis Data NASA POWER dan U.S. Drought Monitor. (deposited 28 Jul 2026 04:22) [Currently Displayed]
Actions (login required)
 |
View Item |