Peramalan Emisi Karbon Menggunakan Metode SARIMA dan LSTM

Abstract

The majority of greenhouse gas (GHG) effects are caused by very high levels of carbon emissions in the world. Therefore, it is necessary to take action to control the levels of carbon emissions in the world. In this study, the world's carbon emission levels were forecasted based on time series data on carbon emissions from 1949 to 2018 in North America. This study uses 2 forecasting methods, namely SARIMA and LSTM, with the consideration that both methods are considered capable of providing good results. Forecasting results show that the best parameter for SARIMA is [(0,1,0) (1,1,0)12] with a MAPE of 1.995%. Meanwhile, if you use the LSTM method with parameters 1 input, 4 hidden layers, and output 1, it produces a MAPE of 0.540%. This condition makes the LSTM method more optimal for predicting carbon emission levels in the world.

The Application of Recurrent Neural Network Method with the Long Short Term Memory (LSTM) Approach to Forecast Hybe Corporation's Stock Price

“…Model ini memiliki kemampuan menyimpan memori yang akan digunakan untuk proses selanjutnya [8] di mana output dari hidden layer akan menjadi input bagi pemrosesan berikutnya [9]. Saat ini banyak peneliti yang mengembangkan pemodelan berbasis neural network salah satunya LSTM yang merupakan bagian dari RNN dengan arsitektur yang sama dengan RNN [10]. Permasalahan yang sering dihadapi dalam menentukan parameter optimal untuk mendapatkan model terbaik dalam machine learning dapat diatasi dengan grid search [11].…”