Parameter: Journal of Statistics

FORECASTING MODEL FOR THE DYNAMICS OF THE CONSUMER PRICE INDEX IN THE HEALTH SECTOR IN EAST JAVA USING THE ARIMA-GARCH MODEL

2025-12-30T11:25:59+00:00

The Consumer Price Index (CPI) is a measure of development success and inflation rates in a country. The CPI is often used to assess the general rate of increase in the prices of goods and services and as a consideration in adjusting salaries, wages, pensions, and other contracts. Therefore, forecasting the CPI is very useful for formulating future policies, including in the health sector. This study aims to create a CPI forecasting model for the health sector in East Java using the ARIMA-GARCH model, namely the integration of the ARIMA model with the GARCH model. The data used were monthly CPI data from January 2020 to December 2023 obtained from the Central Statistics Agency (BPS). The ARIMA model is used to capture long-term trends, while the GARCH model is applied to handle residual heteroscedasticity. The identification results showed that the best ARIMA model is ARIMA(2, 2, 2) with all coefficients statistically significant but heteroscedasticity occurs in the therefore, GARCH modeling is applied to the residuals. Based on the lowest Akaike Information Criterion (AIC) value of 8.342491, the GARCH(1,0) model was selected as the best model. The combined ARIMA(2,2,2)–GARCH(1,0) model produced an AIC value of 18.583 and an RMSE of 0.251383. Residual diagnostic tests indicated that the resulting model meets the assumptions of normal distribution and homogeneity, and there is no significant autocorrelation. The results of this study are expected to contribute to providing predictive information that can be used by the government as a reference in formulating health sector policies, particularly regarding managing the prices of goods and services to maintain economic stability in East Java.

COMPARISON BETWEEN XGBOOST, CATBOOST, RANDOM FOREST, AND LIGHTGBM IN INDONESIAN WOMEN’S BREAST CANCER DATASET

2025-12-30T11:26:00+00:00

Breast cancer is the most prevalent cancer among women in Indonesia and remains a major public health concern, making the identification of key risk factors essential for early detection. This study applies four machine learning classification algorithms—XGBoost, Random Forest, CatBoost, and LightGBM—to classify breast cancer risk factors using a breast cancer dataset consisting of 400 samples. Data preprocessing was performed prior to analysis, and the dataset was divided into 75% training and 25% testing data using 10-fold cross-validation. Model performance was evaluated using accuracy, precision, recall, F1-score, and area under the curve (AUC). The results show that CatBoost outperforms the other models, achieving the highest AUC value of 0.72. Feature importance analysis indicates that a high-fat diet, menopause status, and working status are the most influential risk factors, while breastfeeding shows a protective effect. These findings demonstrate that CatBoost provides strong predictive performance and effectively identifies key factors associated with breast cancer risk in Indonesia.

WATER QUALITY ANALYSIS IN THE RESIDENTIAL AREAS OF PONTIANAK CITY USING THE GEOGRAPHICALLY WEIGHTED LOGISTIC REGRESSION METHOD

2025-12-30T11:26:00+00:00

Water quality is a key indicator of a community’s health and welfare, yet it has deteriorated significantly due to pollution caused by human activities. This study aimed to evaluate Geographically Weighted Logistic Regression’s (GWLR) ability to handle spatial nonstationarity in the relationship between explanatory factors and water quality status in Pontianak City, and to compare its performance with logistic regression. Three modelling approaches were applied to classify water as polluted or non-polluted: (i) logistic regression with spatially invariant) parameters; (ii) GWLR with a fixed Gaussian kernel, producing spatially varying parameters using a fixed bandwidth; and (iii) GWLR with an adaptive Gaussian kernel, producing spatially varying parameters using an adaptive bandwidth. Model performance was compared using Akaike’s Information Criterion (AIC) and classification accuracy. The GWLR model with a fixed Gaussian kernel produced an AIC of 22.52, whereas the logistic regression model produced a slightly lower AIC of 22.39; both models achieved a classification accuracy of 92.86%, with the adaptive-kernel GWLR showing comparable classification performance. These results indicate that, for the parameter settings considered, GWLR offered performance comparable to, but not substantially better than logistic regression for modelling the factors affecting water quality, despite its capacity to address spatial nonstationarity.

MODELING OF LEPROSY CASES IN NORTH SULAWESI USING NEGATIVE BINOMIAL REGRESSION

2025-12-30T11:26:00+00:00

Leprosy is a chronic infectious disease that remains a public health concern in Indonesia, particularly in provinces where the disease is still endemic. North Sulawesi Province is among the regions with relatively high leprosy incidence, indicating the need for province-level analysis to better understand the factors associated with the occurrence of the disease. This study aims to identify the factors influencing the number of leprosy cases in North Sulawesi Province. The explanatory variables considered include the percentage of people living in poverty, the number of health workers, the percentage of toddlers immunized with BCG, the percentage of the population covered by health insurance, the percentage of households with access to clean drinking water, and the percentage of households with access to proper sanitation. The study uses secondary data obtained from official publications of the Central Statistics Agency (BPS) of North Sulawesi Province for the year 2023. Since the response variable is count data, Poisson regression was initially applied. However, due to the presence of overdispersion, Negative Binomial Regression was employed as an alternative modeling approach to obtain more reliable parameter estimates. The results indicate that the percentage of the population covered by health insurance has a statistically significant effect on the number of leprosy cases, with higher coverage associated with a reduction in reported cases. Other variables were found to have no significant effect at the chosen significance level. In conclusion, the findings highlight the importance of health insurance coverage in reducing leprosy incidence in North Sulawesi Province. The use of Negative Binomial Regression proves to be appropriate for modeling overdispersed leprosy case data and can support evidence-based policymaking in leprosy control programs.

THE THE CONTRIBUTION OF VOCATIONAL EDUCATION TO PREDICTING YOUTH UNEMPLOYMENT IN SOUTHEAST SULAWESI: A MACHINE LEARNING APPROACH

2025-12-30T11:26:01+00:00

Youth unemployment remains a major issue in Indonesia, including in Southeast Sulawesi Province. Although the overall open unemployment rate in this province is relatively low, the unemployment rate among young people is still quite high. One contributing factor is the mismatch between educational outcomes and labor market needs, especially for those entering the workforce for the first time. In this context, vocational education is expected to enhance youth employability. Therefore, this study aims to classify youth employment status and identify the predictor that contribute most to the prediction results, particularly vocational education, using SHapley Additive exPlanations (SHAP) values to interpret model decisions. Several machine learning classification methods were evaluated, including naïve Bayes and random forest, with logistic regression used as the baseline comparison model. The findings indicate that the random forest model provides the best classification performance. Based on the analysis, vocational education and age group are the most influential predictors in classifying youth employment status in Southeast Sulawesi Province. Thus, vocational education serves as a key predictor that enhances the model’s ability to classify employment status and is associated with a higher model-predicted probability of being employed

COMPARATIVE ANALYSIS OF UNFUNDED ACTUARIAL LIABILITY BASED ON HULL–WHITE INTEREST RATE ESTIMATION USING ORDINARY LEAST SQUARES AND JACKKNIFE

2025-12-30T11:26:01+00:00

Pension programs are designed to provide financial security after retirement, requiring accurate actuarial valuation to ensure funding adequacy. A key determinant of actuarial liabilities is the interest rate assumption, which directly affects the present value of future pension obligations and the level of unfunded actuarial liability (UAL). Despite its importance, most pension valuation studies rely on deterministic interest rates, while empirical evidence on the use of stochastic interest rate models combined with robust parameter estimation techniques remains limited. This study addresses this gap by evaluating actuarial liability adequacy using the Frozen Initial Liability (FIL) method under a stochastic interest rate framework. The Hull–White one-factor model is employed to capture the dynamic behavior of interest rates, with parameters estimated using Ordinary Least Squares (OLS) and the Jackknife method. The Jackknife approach is introduced to improve estimation robustness, particularly in the presence of small samples and influential observations. Empirical results show that the Jackknife method produces an average interest rate of 0.0678 with a Mean Absolute Percentage Error (MAPE) of 24.4%, while OLS yields an average rate of 0.0665 with a MAPE of 26.1%. Both approaches result in negative UAL values, indicating a fully funded pension scheme with a surplus position. However, the surplus obtained under the Jackknife estimation is lower despite the higher interest rate estimate, suggesting an inverse relationship between interest rates and surplus levels within the FIL framework.

APPLICATION OF THE COPULA FRANK FOR ESTIMATING VALUE AT RISK (VAR) IN TELECOMMUNICATION SUB SECTOR STOCKS

2025-12-30T11:26:01+00:00

Investment is investing capital with the aim of getting money or additional profits. When investing, you need to pay attention to risks that can cause losses for investors. One method that is widely used to measure investment risk is Value at Risk (VaR). VaR often has limitations, especially in capturing non-linear dependencies between variables, so a copula function is needed that can handle moderate to strong dependencies. One of the copulas used is the Archimedian copula with Frank subcopula. This article aims to estimate investment risk using the Value at Risk (VaR) method based on the Frank copula approach and to analyze the dependency structure between stock returns. The main steps in estimating VaR using the Frank copula are calculating the return of each stock, estimating the parameters of the Frank copula, carrying out data simulations using Frank copula parameters, calculating the VaR value using Frank copula. The data used in this research comes from shares of PT. Telkom Indonesia Tbk and shares PT. Indosat Ooredoo Hutchison Tbk. These two stocks have a positive correlation of 0.136. However, such a low correlation may still indicate for non-linear dependencies or tail dependencies that cannot be captured by linear correlations, so additional analysis, namely Frank copula, is required. The estimated Frank copula parameter value is 0.825. From the VaR estimation results, the risk obtained at a 90% confidence level is -0.0222, at a 95% confidence level it is -0.0281 and at a 99% confidence level it is -0.0383.