Perbandingan Model XGBoost Dan Random Forest Untuk Prediksi Tingkat Kemiskinan Di Negara-Negara Anggota OKI

Ferdian, Fadhlan (2026) Perbandingan Model XGBoost Dan Random Forest Untuk Prediksi Tingkat Kemiskinan Di Negara-Negara Anggota OKI. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5002211121-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Kemiskinan merupakan salah satu permasalahan sosial-ekonomi yang masih menjadi perhatian utama di berbagai negara, termasuk negara-negara anggota Organisasi Kerja Sama Islam (OKI). Penelitian ini bertujuan untuk menerapkan dan membandingkan performa algoritma Extreme Gradient Boosting (XGBoost) dan Random Forest dalam memprediksi tingkat kemiskinan berdasarkan indikator makroekonomi. Data yang digunakan merupakan data panel 40 negara anggota OKI pada periode 2007-2021 yang diperoleh dari World Bank, UNDP, dan Legatum Prosperity Index. Variabel dependen dalam penelitian ini adalah tingkat kemiskinan, sedangkan variabel independen meliputi Zakat Rate, Total Population, Unemployment, Domestic General Government Health Expenditure, Inflation, Trade Openness, GDP per Capita, Human Development Index, dan Foreign Direct Investment. Tahapan penelitian meliputi pengumpulan data, pembersihan data, penanganan missing value, log transformation, pembagian data latih dan data uji, pemodelan menggunakan XGBoost dan Random Forest, serta evaluasi performa model. Penelitian ini menggunakan tiga skenario pembagian data, yaitu 60:40, 70:30, dan 80:20, untuk mengetahui konsistensi performa model pada proporsi data latih dan data uji yang berbeda. Evaluasi model dilakukan menggunakan metrik Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), dan R-Squared (R²). Hasil penelitian menunjukkan bahwa model XGBoost menghasilkan performa terbaik pada skenario split 70:30 dengan nilai R² sebesar 0,923, RMSE sebesar 4,763, dan MAE sebesar 3,451. Selain itu, berdasarkan analisis Shapley Additive Explanations (SHAP), variabel GDP_PC menjadi indikator yang paling dominan dalam prediksi tingkat kemiskinan. Setelah variabel GDP_PC dihapus, variabel ZR menjadi indikator paling dominan, diikuti oleh HEALTH_EXP dan HDI. Berdasarkan hasil tersebut, dapat disimpulkan bahwa algoritma XGBoost mampu memberikan performa prediksi terbaik dalam memodelkan tingkat kemiskinan di negara-negara OKI berdasarkan indikator makroekonomi.
=================================================================================================================================
Poverty is one of the socio-economic issues that remains a major concern in many countries, including member countries of the Organisation of Islamic Cooperation (OIC). This study aims to apply and compare the performance of Extreme Gradient Boosting (XGBoost) and Random Forest algorithms in predicting poverty rates based on macroeconomic indicators. The data used in this study are panel data from 40 OIC member countries during the period 2007-2021, obtained from the World Bank, UNDP, and the Legatum Prosperity Index. The dependent variable in this study is the poverty rate, while the independent variables include Zakat Rate, Total Population, Unemployment, Domestic General Government Health Expenditure, Inflation, Trade Openness, GDP per Capita, Human Development Index, and Foreign Direct Investment. The research stages include data collection, data cleaning, handling missing values, log transformation, splitting the data into training and testing sets, model development using XGBoost and Random Forest, and model performance evaluation. This study uses three data-splitting scenarios, namely 60:40, 70:30, and 80:20, to examine the consistency of model performance across different proportions of training and testing data. Model evaluation is carried out using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R-Squared (R²). The results show that the XGBoost model achieves the best performance in the 70:30 split scenario, with an R² value of 0.923, RMSE of 4.763, and MAE of 3.451. Furthermore, based on the Shapley Additive Explanations (SHAP) analysis, GDP_PC is the most dominant indicator in predicting poverty rates. After GDP_PC is removed, ZR becomes the most dominant indicator, followed by HEALTH_EXP and HDI. Based on these results, it can be concluded that the XGBoost algorithm provides the best predictive performance in modeling poverty rates in OIC countries based on macroeconomic indicators.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Kemiskinan, Machine Learning, OIC Countries, Panel Data, Random Forest, XGBoost, Machine Learning, OIC Countries, Panel Data, Poverty, Random Forest, XGBoost
Subjects:	H Social Sciences > HA Statistics H Social Sciences > HB Economic Theory Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.9.D343 Data mining. Querying (Computer science)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Fadhlan Ferdian
Date Deposited:	03 Aug 2026 00:54
Last Modified:	03 Aug 2026 00:54
URI:	http://repository.its.ac.id/id/eprint/140597

Actions (login required)

View Item