Perancangan Mobile Predictor Cuaca Maritim Menggunakan Metode Hybrid Logika Fuzzy Tipe 2-Jaringan Saraf Tiruan Dengan Optimasi Algoritma Differential Evolution

Kurniawan, Muhammad Rifki (2018) Perancangan Mobile Predictor Cuaca Maritim Menggunakan Metode Hybrid Logika Fuzzy Tipe 2-Jaringan Saraf Tiruan Dengan Optimasi Algoritma Differential Evolution. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.

[img]
Preview
Text
02311440000034-Undergraduate_Theses.pdf - Accepted Version

Download (8MB) | Preview

Abstract

Informasi cuaca maritim baik saat ini dan yang akan datang (prediksi) sangat berguna sebagai sistem pendukung dalam berbagai aktivitas maritim. Informasi tersebut sangat penting bagi nelayan untuk menentukan kelayakan jalur pelayaran. Bagi industri, informasi tersebut dapat digunakan sebagai pertimbangan aktivitas maritim seperti pada pembangkit energi terbarukan. Dalam konteks prediksi, beberapa prediktor masih menggunakan pendekatan statistik, di mana pendekatan ini memiliki akurasi rendah dan tidak dapat bekerja pada data nonliner. Penelitian ini merancang prediksi cuaca maritim menggunakan pendekatan artificial intelligence berupa sistem prediktor hybrid. Prediktor ini menggunakan prinsip persilangan antara dua metodologi, yaitu Interval Type 2 Fuzzy Inference System (IT2FIS) dan Jaringan Saraf Tiruan-Propagasi Balik (JST-BP). Di mana, prediktor hybrid bekerja dengan memberikan koefisien pada masing-masing keluaran prediktor. Koefisien-koefisien tersebut ditentukan berdasarkan proses optimasi dengan algoritma Differential Evolutin (DE). Hasil prediksi tersebut ditampilkan kepada pengguna melalui User Interface berbasis Android, sehingga informasi cuaca dapat diakses menggunakan mobile smartphone di manapun dan kapanpun oleh siapapun secara realtime. Penelitian ini menemukan bahwa prediktor hybrid memiliki keunggulan performa baik dalam parameter RMSE maupun MAPE pada semua variabel dibandingkan prediktor tunggal masing-masing (JST dan IT2FIS) saat proses pelatihan. Namun, dalam tahap pengujian performa prediktor hybrid tidak selalu lebih baik dari keduanya. Berdasarkan standar prediktor berbasis MAPE, prediktor suhu dan kelembaban termasuk ke dalam prediktor highly accurate, kecepatan angin termasuk reasonable forecast, dan prediktor arah angin dan curah hujan termasuk ke dalam inacurate forecast. Sementara, percobaan prediksi beberapa hari ke depan menghasilkan pola prediksi yang semakin konvergen seiring bertambahnya hari ke depan yang diprediksi, yang mana pola ini berlaku pada semua variabel. ============================================================ Current and future maritime weather information (predictions) is very useful as a support system in various maritime activities. Such information is very important for fishermen to determine the feasibility of shipping lanes. For the industry, the information can be used as consideration of maritime activities such as in renewable energy generation. In the context of prediction, some predictors still use statistical approaches, where this approach has low accuracy and can not work on nonliner data. This research designed maritime weather prediction used artificial intelligence approach in the form of hybrid predictor system. This predictor used the principle of crossing between two methodologies, namely Interval Type 2 Fuzzy Inference System (IT2FIS) and Artificial Neural Network-Back Propagation (JST-BP). Where, hybrid predictors worked by giving coefficients to each predictor output. The coefficients were determined based on the optimization process with Differential Evolutin (DE) algorithm. The prediction results were displayed to users via Android-based User Interface, so weather information can be accessed using mobile smartphones wherever and whenever by anyone in realtime. This study found that hybrid predictors have superior performance in both RMSE and MAPE parameters on all variables compared to their respective single predictors (ANN and IT2FIS) during the training process. However, in the testing phase hybrid predictor performance was not always better than both. Based on MAPE-based predictor standards, temperature and humidity predictors were included in the highly accurate predictor, wind speed including reasonable forecast, and predictors of wind direction and rainfall included in forecast forecast. Meanwhile, predictive experiments over the next few days resulted in an increasingly convergent prediction pattern with predicted days ahead, which this pattern applies to all variables.

Item Type: Thesis (Undergraduate)
Uncontrolled Keywords: hybrid, jaringan saraf tiruan, inteval type 2 fuzzy inference system, differential evolution, cuaca maritim, mobile.
Subjects: Q Science > QA Mathematics > QA336 Artificial Intelligence
Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Q Science > QA Mathematics > QA9.58 Algorithms
Q Science > QA Mathematics > QA9.64 Fuzzy logic
Q Science > QC Physics > QC866.5 Climatology--Forecasting.
T Technology > T Technology (General) > T174 Technological forecasting
Divisions: Faculty of Industrial Technology > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User: Muhammad Rifki Kurniawan
Date Deposited: 01 Jul 2021 06:59
Last Modified: 01 Jul 2021 06:59
URI: https://repository.its.ac.id/id/eprint/56571

Actions (login required)

View Item View Item