Rancangan Arsitektur dalam Peramalan Curah Hujan Menggunakan Metode Long Short - Term Memory dan Artificial Neural Network

Purwanto, Aruni Rahmaniar (2024) Rancangan Arsitektur dalam Peramalan Curah Hujan Menggunakan Metode Long Short - Term Memory dan Artificial Neural Network. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6003211002-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 March 2026. Download (2MB) | Request a copy

Abstract

Indonesia merupakan suatu negara yang memiliki dua musim yaitu, musim kemarau dan musim hujan. Hujan sangat erat kaitannya dengan sektor pertanian. Curah hujan yang tinggi akan menyebabkan dampak negatif terhadap masyarakat. Curah hujan yang tinggi dapat menyebabkan gagal panen yang disebabkan oleh banyaknya kandungan air pada area persawahan. Agar dapat meminimalisir dampak kerugian yang terjadi, penelitian ini ingin meramalkan curah hujan Kabupaten Jember. Data curah hujan harian merupakan suatu data yang memiliki banyak nilai nol atau disebut data intermittent. Data curah hujan merupakan data yang cukup sulit untuk dimodelkan secara linear, sehingga pada penelitian ini dilakukan model Self Exciting Threshold Autoregressive (SETAR) pada metode Long Short – Term Memory (LSTM). Metode long short – term memory merupakan salah satu pengembangan dari metode artificial neural network yang biasa digunakan pada kasus peramalan. Dalam melakukan peramalan, dibutuhkan arsitektur yang sesuai agar model yang dihasilkan baik. Oleh karena itu, dilakukan beberapa rancangan arsitektur LSTM yang berbeda untuk dapat melihat performa peramalan menggunakan metode LSTM yang terbaik. Arsitektur LSTM yang menghasilkan RMSE paling rendah didapat melalui model dengan fungsi aktivasi sigmoid (1 hidden layer, 25 neuron hidden layer). Hasil peramalan menggunakan rancangan arsitektur terbaik metode long short – term memory dan metode SETAR – LSTM akan dibandingkan dengan hasil peramalan menggunakan metode artificial neural network. Pemodelan SETAR dilakukan dengan menggunakan 2 regime sehingga model yang dihasilkan SETAR(2,1,2). Ketiga metode yaitu, LSTM, SETAR – LSTM, ANN – Backpropagation memiliki perbedaan nilai akurasi. Akurasi terbaik diperoleh melalui metode LSTM.
====================================================================================================================================
Indonesia is a country that has two seasons, namely, the dry season and the rainy season. Rain is closely related to the agricultural sector. High rainfall will cause negative impacts on the community. High rainfall can cause harvest failure caused by the large amount of water content in paddy fields. In order to minimize the impact of losses that occur, this study wants to forecast the rainfall of Jember Regency. Daily rainfall data is a data that has many zero values or called intermittent data. Rainfall data is quite difficult data to model linearly, so in this study the Self Exciting Threshold Autoregressive (SETAR) model was carried out on the Long Short – term Memory (LSTM) method. The LSTM method is one of the developments of the artificial neural network method commonly used in forecasting cases. In forecasting, an appropriate architecture is needed so that the resulting model is good. So that several different LSTM architectural designs were carried out to be able to see forecasting performance using the best LSTM method. The LSTM architecture that produces the lowest RMSE is obtained through a model with sigmoid activation functions (1 hidden layers, 25 hidden layer neurons). The forecasting results using the best architectural design, the LSTM method and the SETAR-LSTM method will be compared with the forecasting results using the artificial neural network method. SETAR modeling is carried out using 2 regimes so that the resulting model is SETAR(2,1,2). The three methods namely, LSTM, SETAR – LSTM, ANN – Backpropagation have different accuracy values. The best accuracy was obtained through the LSTM method.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	akurasi, backpropagation, long short - term memory, SETAR, accuracy
Subjects:	Q Science > Q Science (General) > Q325.78 Back propagation Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49101-(S2) Master Thesis
Depositing User:	Aruni Rahmaniar Purwanto
Date Deposited:	19 Feb 2024 08:00
Last Modified:	19 Feb 2024 08:00
URI:	http://repository.its.ac.id/id/eprint/107602

Actions (login required)

View Item