Efektivitas Metode Hibrida ARIMA-MLP untuk Peramalan Nilai Tukar Petani
The agricultural sector remains a crucial pillar of Indonesia’s economy, making the most significant contribution. Still, the situation of farmers, primarily the elderly, indicates physical limitations and low income leading to high poverty levels, coupled with fluctuations in the Farmer Exchange Ra...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Department of Mathematics, Universitas Negeri Gorontalo
2024-02-01
|
| Series: | Jambura Journal of Mathematics |
| Subjects: | |
| Online Access: | https://ejurnal.ung.ac.id/index.php/jjom/article/view/23944 |
| Summary: | The agricultural sector remains a crucial pillar of Indonesia’s economy, making the most significant contribution. Still, the situation of farmers, primarily the elderly, indicates physical limitations and low income leading to high poverty levels, coupled with fluctuations in the Farmer Exchange Rate (FER) annually tending to decline in D.I. Yogyakarta, indicating losses due to increased production costs. This research aims to assess the effectiveness of the Hybrid Autoregressive Integrated Moving Average (ARIMA) – Multilayer Perceptron (MLP) method in forecasting NTP in D.I. Yogyakarta. This is based on the analysis of comparing the accuracy values of forecasts using Mean Absolute Percentage Error (MAPE) evaluation or through visualizing the forecast graphs generated between the ARIMA and Hybrid ARIMA-MLP methods. The combination (hybrid) of ARIMA and MLP methods addresses the complexity of time series, where ARIMA anticipates NTP changes by handling linear patterns. At the same time, MLP improves forecast accuracy by managing more complex patterns (both linear and nonlinear). Thus, it can provide more accurate information about the welfare development of farmers. The results show that the Hybrid ARIMA-MLP method is significantly better than the individual ARIMA method, with the obtained model being Hybrid ARIMA-MLP (12-5-10-2) and an accuracy of 99.993%. |
|---|---|
| ISSN: | 2654-5616 2656-1344 |