cover
Article Per Year (5 Year)
All
Home Page
OAI Link
Editorial Team
Contact
Reviewer
Google Scholar
Contact Name
Nusar Hajarisman
Contact Email
statistika@unisba.ac.id
Phone
+628124569091
Journal Mail Official
statistika@unisba.ac.id
Editorial Address
Program Studi Statistika Universitas Islam Bandung Jl. Ranggagading No. 8 Bandung 40116 Phone: 081572198581 Email: nusarhajarisman@unisba.ac.id
Location
Kota bandung,
Jawa barat
INDONESIA
STATISTIKA
Published by Universitas Islam Bandung
ISSN : 14115891     EISSN : 25992538     DOI : https://doi.org/10.29313/statistika.v23i1.1628
Core Subject : Science, Education,
Decision Sciences, Operations Research & Management Mathematics
STATISTIKA published by Department of Statistics, Faculty of Mathematics and Natural Sciences, Bandung Islamic University as pouring media and discussion of scientific papers in the field of statistical science and its applications, both in the form of research results, discussion of theory, methodology, computing, and review books. Published biannually in May and November each.
Arjuna Subject : -
Articles 99 Documents
 6   7   8   9   10 
LASSO-Regularized Binary Logistic Regression on Imbalanced Mode Choice Data Eko Primadi Hendri; Novi Urbaningrum; Sarah Fadhlia
Statistika Vol. 25 No. 2 (2025): Statistika
Publisher : Department of Statistics, Faculty of Mathematics and Natural Sciences, Universitas Islam Bandung

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.29313/statistika.v25i2.8126

Abstract

Abstract. Binary logistic regression is a widely used method for modeling mode choice, but it often suffers from reduced predictive accuracy when dealing with high-dimensional datasets and class imbalance. This study implements binary logistic regression with LASSO regularization to identify significant factors influencing transportation mode choice between motorcycles and Trans Metro buses in the CBD of Pekanbaru. Data from 100 respondents were collected through revealed-preference and stated-preference surveys, with class imbalance (71% motorcycle, 29% bus) addressed using the Synthetic Minority Over-sampling Technique (SMOTE). Model performance was evaluated using accuracy, AUC, precision, recall, and F1-score via Repeated Random Subsampling Validation (RRSV). Results show that the LASSO model with SMOTE increased recall from 0.125 to 0.25 and F1-score from 0.143 to 0.267 compared to the non-SMOTE model, with an accuracy of 0.621 and an AUC of 0.613, indicating improved ability to detect the minority class. Statistically significant predictors include occupation, monthly income, and ownership of an alternative vehicle. This study demonstrates that combining LASSO and SMOTE is effective in handling imbalanced data, providing strong quantitative evidence to support urban transport policy planning.
Clustering of Stunting Determinants in Papua Province Using Average Linkage Method and K-Means Algorithm Dina Jumiatul Fitri; Nicea Roona Paranoan; Caecilia Bintang Girik Allo; Tiku Tandiangnga; Winda Ade Fitriya B.
Statistika Vol. 26 No. 1 (2026): Statistika
Publisher : Department of Statistics, Faculty of Mathematics and Natural Sciences, Universitas Islam Bandung

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.29313/statistika.v26i1.8819

Abstract

Abstract. In the current public health context, stunting remains a chronic nutritional problem that requires serious attention in Indonesia, especially in the eastern region such as Papua. This study aims to cluster districts and cities in Papua Province based on factors influencing stunting prevalence using both hierarchical and non-hierarchical clustering approaches, namely the Average Linkage method and the K-Means algorithm. The study acquires secondary data from the 2023 Indonesia Health Survey and the 2024 Statistics Indonesia (BPS) reports, covering nine districts and cities in Papua Province. The variables used include poverty rate, access to proper sanitation, access to safe drinking water, number of midwives, and number of community health centers (puskesmas). The hierarchical clustering with Average Linkage produced two clusters with a Silhouette Coefficient value of 0.239. In comparison, the K-Means method, with the optimal number of clusters determined by the Elbow method and Silhouette analysis, formed three clusters and achieved a higher Silhouette Coefficient value of 0.32, indicating better cluster compactness and separation. The K-Means results provide a more detailed segmentation, distinguishing areas with basic infrastructure inequality, regions with social vulnerability and limited services, and districts with relatively advanced service access and lower poverty. Overall, the K-Means approach offers stronger clustering performance and more granular regional classification. Therefore, clustering results can serve as an evidence-based foundation for designing more targeted regional development and health intervention policies aimed at reducing stunting prevalence in high-risk areas of Papua.
Interpretable Tree-Based Models for Identifying Youth Not in Employment, Education, or Training Siti Nur Azizah; Nabila Syukri; Mauizatun Hasanah; Nimas Ayu Hapsari; Bagus Sartono; Aulia Rizki Firdawanti; Budi Susetyo; Gerry Alfa Dito
Statistika Vol. 26 No. 1 (2026): Statistika
Publisher : Department of Statistics, Faculty of Mathematics and Natural Sciences, Universitas Islam Bandung

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.29313/statistika.v26i1.9117

Abstract

Abstract. Youth Not in Employment, Education, or Training (NEET) is an important indicator in evaluating the extent to which Indonesia's demographic bonus can be optimally utilized. West Java, as the province with the largest number of young people and a NEET rate above the national average, is a strategic context for analysis. Using large-scale SUSENAS 2024 data with class imbalance, this study aims to: (1) compare the performance of Decision Tree, Random Forest, and XGBoost in predicting NEET status under various class imbalance handling strategies; (2) identify key features contributing to predictions using SHAP; and (3) identify NEET typologies based on feature contribution similarities. The Random Forest–Class Weight model showed the best performance with a balanced accuracy of 0.7519. SHAP analysis identified eight key predictors contributing to NEET status, namely age, highest level of education, per capita expenditure, savings account ownership, access to financial services, marital status, KIP/PIP receipt, and gender. Local SHAP-value-based clustering yielded five NEET profiles with distinct combinations of risk factors. Two clusters showed the most distinct risk patterns, each characterized by economic vulnerability and limitations due to domestic roles, making them most easily recognized by the model, while the other clusters showed more complex risk patterns. These findings indicate that NEET status is formed through various risk pathways influenced by economic, educational, and demographic factors, thus requiring the formulation of more targeted policies.
Default Risk Modeling in Credit Insurance with Generalized Non-Linear Model Zakaria Bani Ikhtiyar; Lathifatul Aulia; Wulan Bhakti Pertiwi; Muhammad Sulthan Madany; Nova Putri Ardelia; Daffa Haidar Maulana
Statistika Vol. 26 No. 1 (2026): Statistika
Publisher : Department of Statistics, Faculty of Mathematics and Natural Sciences, Universitas Islam Bandung

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.29313/statistika.v26i1.9208

Abstract

Abstract. The growth of credit insurance in Indonesia has increased insurers’ exposure to default risk, particularly defaults arising from debtor mortality. In practice, mortality-related default risk is commonly assessed using static life tables or deterministic assumptions, which fail to capture stochastic mortality dynamics and heterogeneity among debtors. This limitation often leads to inaccuracies in estimating default probabilities and expected claims, potentially compromising pricing and reserving decisions. This study addresses this gap by developing an integrated framework that combines stochastic mortality estimation with credit default risk modeling. The primary contribution of this research is the synthesis of the PLAT stochastic mortality model with a Generalized Non-Linear Model (GNLM). The PLAT model is employed to capture age, period, and cohort effects, while the GNLM integrates these estimated mortality probabilities with loan-specific characteristics, such as loan amount, tenor, and underwriting year. Furthermore, mortality rates are estimated within a Quasi-Poisson regression framework to account for overdispersion in claim data. Our results indicate that default probabilities significantly increase with age and vary across underwriting cohorts, demonstrating that loan characteristics materially influence the magnitude of mortality-driven risk. This proposed framework provides a more robust actuarial basis for pricing, reserving, and risk management in credit insurance.
ARIMA-GARCH Based EWMA Control Chart for Rupiah-US Dollar Exchange Rate Abrar Usman; Erna Tri Herdiani
Statistika Vol. 26 No. 1 (2026): Statistika
Publisher : Department of Statistics, Faculty of Mathematics and Natural Sciences, Universitas Islam Bandung

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.29313/statistika.v26i1.9687

Abstract

Abstract. The Exponentially Weighted Moving Average (EWMA) control chart is more responsive to small shifts because it assigns greater weight to recent observations. This property is relevant for monitoring the Indonesian Rupiah-U.S. Dollar (IDR/USD) exchange rate, which is highly sensitive to global dynamics and exhibits substantial fluctuations. However, exchange-rate time series often exhibit serial correlation and conditional heteroskedasticity, which can reduce the accuracy of conventional control charts. To address this issue, an Autoregressive Integrated Moving Average-Generalized Autoregressive Conditional Heteroskedasticity (ARIMA-GARCH) model provides a more reliable statistical basis by capturing both temporal dependence and volatility dynamics. This study aims to develop an EWMA control chart based on ARIMA-GARCH residuals for monitoring the IDR/USD exchange rate and to evaluate its performance using the in-control average run length (IC-ARL) and out-of-control average run length (OC-ARL). The best ARIMA-GARCH specification was selected using the Akaike information criterion and parameter significance tests. Residuals from the selected model, which were free from serial correlation and remaining heteroskedasticity, were used to construct EWMA charts with smoothing parameters  = 0.05, 0.1, and 0.3 and control limits  = 3. The ARIMA (3,1,2)-GARCH (1,1) model was identified as the best specification. Based on its residuals, the EWMA chart with  = 0.1 detected 20 out-of-control points associated with periods of global instability while maintaining a low false-alarm rate. Overall, the proposed ARIMA-GARCH-based EWMA chart improves monitoring sensitivity and stability compared with a conventional EWMA chart, and  = 0.1 provides the most balanced performance.
National Electricity Demand Forecasting using Hybrid Statistical-Scenario and Monte Carlo Simulation Dira Dini Dian Kemala
Statistika Vol. 26 No. 1 (2026): Statistika
Publisher : Department of Statistics, Faculty of Mathematics and Natural Sciences, Universitas Islam Bandung

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.29313/statistika.v26i1.9776

Abstract

Abstract. National electricity demand is a key strategic indicator for planning and managing of Indonesia’s electricity system. Economic growth, demographic dynamics, and energy policies drive a steady increase in electricity demand, creating uncertainty in forecasting electricity demand. This study aims to develop an electricity demand forecasting model based on a hybrid statistical approach that probabilistically models uncertainty. The Autoregressive Integrated Moving Average with Exogenous Variables (ARIMAX) model is used to capture the influence of exogenous variables, then combined with Monte Carlo simulations to produce scenario-based projections. Secondary data from BPS, PLN, ESDM, and the Ministry of Trade, covering for the period 2003–2024 are used in the analysis. Based on evaluations of several performance metrics, the ARIMAX (1,1,1) model was selected as the best. The electrification ratio variable is modeled using a beta distribution, while the GDP growth, population, and electricity tariff variables are modeled using lognormal distributions. The projection results for the period 2025–2035 show that all scenarios produce electricity consumption ranging from 381 to 430 terawatt-hours (TWh) in 2035. Monte Carlo simulations suggest that the electrification ratio primarily affects projection uncertainty, while economic and demographic variables remain relatively stable. This approach provides more comprehensive probabilistic range information than a single deterministic estimate, thereby supporting more adaptive, risk-based energy planning.
Analysis of Factors Influencing QRIS Usage in Banten, DKI Jakarta, and West Java Provinces Yekti Widyaningsih; Galih Nur Kantaatmaja
Statistika Vol. 26 No. 1 (2026): Statistika
Publisher : Department of Statistics, Faculty of Mathematics and Natural Sciences, Universitas Islam Bandung

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.29313/statistika.v26i1.9826

Abstract

Abstract. Quick Response Code Indonesian Standard is a non-cash payment method developed to support the digital transformation of Indonesia’s national payment system. Through a single QR code, various digital wallet applications and mobile banking platforms can be interconnected, offering convenience in transactions for both businesses and consumers. Although various factors, such as demographic differences and user behavior, are presumed to influence QRIS usage, there has been limited research exploring these aspects in depth. Therefore, this study aims to analyze the factors that influence QRIS usage by employing the ordinal logistic regression model, focusing on both demographic attributes and smartphone usage behavior. The ordinal logistic regression is a statistical technique designed to examine the relationship between an ordinal response variable and one or more independent variables. This model is suitable for analyzing data where the dependent variable consists of ordered categories. Overall, income and smartphone type emerge as consistently significant variables across all studied regions. Users with higher income levels and those using iOS-based smartphones tend to exhibit higher levels of QRIS usage compared to other groups. Based on the k-fold cross-validation method, the ordinal logistic regression model employed in this study yields an overall accuracy rate of 65.22%. These results indicate that both demographic factors and smartphone usage behavior have a significant impact on QRIS adoption in the observed regions.
Loan Estimation with Aggregate Distribution Non Homogeneous Poisson Process (NHPP) – Johnson SB Adri Arisena; Espreilla Harin Widhastika; Achmad Zanbar Soleh
Statistika Vol. 26 No. 1 (2026): Statistika
Publisher : Department of Statistics, Faculty of Mathematics and Natural Sciences, Universitas Islam Bandung

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.29313/statistika.v26i1.9970

Abstract

Abstract. The Consumer Cooperative of IKOPIN University (IU-Coop) operates several business units, with the Savings and Loan Unit (USP) contributing the most to the cooperative’s financial performance. However, unpredictable loan requests and reliance on member deposits have led to cash flow instability. To address this issue, this study proposes a predictive model for loan demand using the Non-Homogeneous Poisson Process (NHPP), which effectively models random events with time-varying intensity. Additionally, the Johnson SB distribution is employed to model loan amounts constrained within a specific range, while the Generalized Pareto distribution is used to represent the distribution of disbursement times, capturing the possibility of extreme delays. The dataset comprises short-term loan records from January 2022 to October 2024, with variables including loan amount (X) and disbursement duration (N). The analysis reveals that borrowers in Group 1 (loans under IDR 5,000,000) are expected to request two disbursements by the end of December 2024, requiring a total fund preparation of IDR 5,598,828. This modeling approach enhances liquidity management by providing more accurate forecasts and supports data-driven decision-making in microfinance operations.
Time Series Clustering and Mean–Variance Portfolio Modeling on IDX Sharia Growth Stocks Fitri Amanah; Fauziah Roshafara; Nafa Nurhanifah; Novianda Dwi Tanti Ramdani
Statistika Vol. 26 No. 1 (2026): Statistika
Publisher : Department of Statistics, Faculty of Mathematics and Natural Sciences, Universitas Islam Bandung

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.29313/statistika.v26i1.9723

Abstract

Abstract. A stock portfolio represents an investment strategy aimed at maximizing returns while minimizing risk. Stock data, as a form of time series data, relies on historical information to evaluate performance and identify price movement patterns. Consequently, stock selection is a critical component in asset diversification. The novelty of this study is to apply time series clustering methods to stock data and utilize the resulting clusters to construct a mean-variance portfolio. The data used are the weekly closing prices of 10 Sharia-Growth Index stocks for the period December 2022 to November 2024. Using Dynamic Time Warping (DTW) distance with average linkage, two clusters were identified: cluster 1 (MPMX, TLKM) and cluster 2 (MAPI, HEAL, ISAT, AKRA, BMTR, SIDO, KLBF, PWON), with a Silhouette coefficient of 0.9471 indicating strong clustering performance. To construct the mean-variance portfolio, three stocks with positive expected returns were selected: HEAL, ISAT, and MAPI, which are members of Cluster 2. The optimal weights obtained using Lagrange optimization are HEAL (0.49%), ISAT (99.51%), and MAPI (0%). The Lagrange method allocates 0% to MAPI because its risk level (variance of 0.00322) is the highest among the three candidate stocks, thereby not helping to minimize the overall portfolio risk. Therefore, the optimal portfolio formed by combining HEAL and ISAT would have provided an estimated return of 0.56% and a risk of 4.1%.

Page 10 of 10 | Total Record : 99

 6   7   8   9   10 

Filter by Year

2021 2026


Reset
Filter By Issues
All Issue Vol. 26 No. 1 (2026): Statistika Vol. 25 No. 2 (2025): Statistika Vol. 25 No. 1 (2025): Statistika Vol. 24 No. 2 (2024): Statistika Vol. 24 No. 1 (2024): Statistika Vol. 23 No. 2 (2023): Statistika Vol. 23 No. 1 (2023): Statistika Vol. 22 No. 2 (2022): Statistika Vol. 22 No. 1 (2022): Statistika Vol. 21 No. 2 (2021): Statistika