Peramalan Jumlah Impor Gula Menggunakan Metode Panel Data dan Machine Learning

Mas, Syahdan Filsafan (2024) Peramalan Jumlah Impor Gula Menggunakan Metode Panel Data dan Machine Learning. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text (Full Signed) final tesis mas syahdan filsafan (1).pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (1MB) | Request a copy

Abstract

Accurate forecasting of sugar demand is important to deal with market fluctuations. Factors such as consumption trends, consumer preferences, and economic conditions affect demand. Previous research shows that production, consumption, domestic and international prices, exchange rates, GDP, and import tariffs have a significant effect on sugar imports in Indonesia.
This research utilizes Machine Learning and Deep Learning methods to update previous studies. Panel data is used to capture the heterogeneity of observations, overcoming the limitations of Time Series data. K-Nearest Neighbors (KNN) Imputer algorithm with K-3 is used to overcome missing values.
The methods used include Deep Learning (LSTM) and Classical Machine Learning, which are applied to Time Series data for forecasting. The results will be compared with Panel Data Regression Analysis. Data is sourced from the Badan Pusat Statistik (BPS). The evaluation uses RMSE, precision, recall, and f1-score.
A series of tests were conducted to determine the best model. KNN with K=3 proved to be the most effective in handling missing values. Fixed Effects Model (FEM) was selected for panel data regression through a series of statistical tests.
Results show MLP Regressor performs best with the highest R2 (0.6456), outperforming other models such as LightGBM, GradientBoosting, LSTM, Gated Recurrent Unit (GRU), and Convolutional Neural Network (CNN) in predicting sugar imports in Indonesia. This finding suggests that for this dataset, MLP models tend to perform better than other models.

Item Type:	Thesis (Masters)
Subjects:	T Technology > T Technology (General) > T174 Technological forecasting
Divisions:	Faculty of Industrial Technology > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User:	Mas Syahdan Filsafan
Date Deposited:	07 Aug 2024 08:20
Last Modified:	07 Aug 2024 08:20
URI:	http://repository.its.ac.id/id/eprint/113519

Actions (login required)

View Item