FORSTAT International Conference

The Effect of Socioeconomic Indicators on Environmental Quality in Indonesia Using Logistic Regression

Melani Yusi Aryanda — 2026-06-23

Environmental degradation in Indonesia is becoming increasingly evident as development activities and the exploitation of natural resources intensify. Deforestation resulting from the expansion of plantations, mining, and industrial development leads to the loss of forest cover as well as a decline in air quality and biodiversity. These conditions indicate that rapid economic growth often places significant pressure on environmental sustainability if not balanced by sustainable management. This study aims to analyze the relationship between socioeconomic factors and the risk of environmental degradation in Indonesia in 2024. The study employs a logistic regression model using secondary data from 38 provinces sourced from the Central Statistics Agency (BPS) and the Ministry of Environment and Forestry (KLHK). Independent variables include per capita GRDP, poverty rate, Human Development Index, and population density, while the dependent variable is environmental quality measured through the Environmental Quality Index classified based on the national median. The analysis results indicate that the model is statistically significant simultaneously and can moderately explain variations in environmental quality. Partially, the poverty rate variable shows a significant negative effect on environmental quality at the 10% significance level. Meanwhile, the per capita GRDP, Human Development Index, and population density variables do not show a significant effect. These findings indicate that the risk of environmental quality decline is not solely determined by the level of prosperity but is influenced by structural dynamics and inter-regional development pressures.

Extreme Value Theory-Based Value at Risk for Stock Portfolio Risk Estimation: A Comparative Study of GEV and GPD Models in Indonesian Banking Stocks

Umar Sodiq — 2026-06-23

Risk management is a crucial element in stock investment activities to minimize the negative impacts of market uncertainty, particularly during periods of extreme price fluctuations. One effective statistical approach for measuring maximum potential loss in return distributions characterized by fat tails is Value at Risk (VaR) integrated with Extreme Value Theory (EVT). This study aims to analyze stock portfolio risk, estimate potential losses, and evaluate the comparative accuracy of the Block Maxima and Peak Over Threshold methods. The Block Maxima method models maximum extreme values within specific time blocks using the Generalized Extreme Value (GEV) distribution, whereas the Peak Over Threshold method models data exceeding a predefined threshold (u) using the Generalized Pareto Distribution (GPD). The research sample consists of five banking subsector stocks with the largest market capitalization, BBCA, BBRI, BMRI, BBNI, and BRIS for the period from May 1, 2019, to May 31, 2025. Results indicate that VaR estimation at a 95% confidence level using the VaR-GEV method yields a potential loss of 6.64%, while the VaR-GPD method yields 3.13%. Evaluation via the Kupiec backtesting test reveals that the VaR-GEV model is invalid due to being overly conservative with an excessively low violation rate. Conversely, the VaR-GPD model is proven valid, with an actual violation rate of 5.49% that closely aligns with expected values. Given its superior validity and accuracy, the Value at Risk Generalized Pareto Distribution approach is determined to be the optimal model for estimating market risk under extreme conditions.

Comparison of GWANN and GWNBR in the Analysis of Positive Covid-19 Factors

Risma Tri Rahmawati — 2026-06-23

Covid-19 emerged with a rapid transmission rate, yet the precise factors contributing to its spread were initially unclear. To reduce the number of positive Covid-19 cases in East Java, spatial analysis is essential for examining these influencing factors. This study employs local research approaches, utilizing adaptive bisquare kernel weights. The local research, where the estimated value differs by region, is carried out using the Geographically Weighted Negative Binomial Regression (GWNBR) and the Geographically Weighted Artificial Neural Network (GWANN) method. This research aims to compare the GWANN, and GWNBR methods to determine the more effective approach for analyzing Covid-19 factors in East Java. The best model is selected based on the R²and RMSE values. The findings indicate that the Geographically Weighted Negative Binomial Regression (GWNBR) method is the more effective model, with an RMSE of 936.2055 and an R²of 0.902, and subsequent clustering based on significant local coefficients produced four regional groups; notably, comorbidity prevalence was significant in all areas, while other determinants differed by locality. These findings indicate that accounting jointly for overdispersion and spatially varying relationships improves inference on Covid-19 case patterns and highlights comorbidity burden as a province-wide priority for intervention.

Modeling the Categorical Patterns of Maggot Cultivation Knowledge Based on Correspondence Analysis in the Balangan Environmental Awareness Group in Kalurahan Wukirsari

Avorey Bias Agung Valentino Duha — 2026-06-23

Organic waste management remains a major environmental challenge in Kalurahan Wukirsari, Sleman Regency, Special Region of Yogyakarta Province, Indonesia. The cultivation of Black Soldier Fly (BSF) maggots has been implemented by Environmental Awareness Group as a sustainable solution. The success of this program is closely related to the knowledge level of its members. This study aims to model the categorical patterns of maggot cultivation knowledge using Correspondence Analysis and to identify the dominant influencing factors. Primary data were collected through questionnaires from 50 group members. The analysis included descriptive statistics, validity and reliability tests, chi-square tests, and correspondence analysis. The results show that most respondents have good to very good knowledge levels. Significant relationships were found between knowledge level and respondents characteristics, especially age, education, and occupation (p-value < 0.05). Correspondence analysis indicates that younger respondents (≤35 years) tend to have lower knowledge, while older respondents (>55 years) are associated with very good knowledge. Higher education and certain occupations, such as village officials and employees, are also linked to higher knowledge levels. The mapping results provide a clear basis for designing targeted training and empowerment programs.

Spatial and Temporal Dynamics of Measles Cases in DKI Jakarta, Indonesia, 2016 to 2025

Muhamad Nobel Fauzan — 2026-06-23

Measles remains a significant public health threat in densely populated urban settings like DKI Jakarta, Indonesia, where sustaining herd immunity is challenging. To optimize public health responses, the main objective of the research is to evaluate the spatial and temporal dynamics of measles over a ten-year period from January 2016 to December 2025 using Local Indicators of Spatial Association (LISA). Epidemiological data were obtained from the Jakarta Provincial Health Office, aggregating case reports exclusively from hospitals across the province. The analysis involved calculating incidence rates and applying spatial clustering techniques to detect localized transmission hotspots. Results indicate a total of 6,951 cases, revealing a cyclical three-year outbreak pattern with significant peaks in 2017, 2023, and an unprecedented surge in 2025 totaling 4,616 cases. Spatial analysis identified persistent high-risk transmission hotspots in North Jakarta, specifically Koja and Cilincing, alongside a newly emerging hotspot in Kalideres, West Jakarta. Furthermore, children under five years old experienced the highest incidence rates, particularly within these specific hotspot districts. The conclusion emphasizes that prioritizing localized interventions and targeted vaccination strategies in these defined geographic clusters and vulnerable age demographics is essential to control transmission and mitigate future outbreaks.

Forecasting Wholesale Rice Prices in Indonesia Using an ARIMAX Model with ENSO and Hydrometeorological Disaster Variables

Nadya Ilma Pratiwi — 2026-06-23

Rice is the main staple food in Indonesia, making its price stability a critical component of national food security. However, rice prices often fluctuate due to various factors, including climate variability and hydrometeorological disasters. One of the global climate drivers affecting agricultural conditions is the El Niño–Southern Oscillation (ENSO), which influences rainfall patterns and crop productivity. In addition, disasters such as floods and droughts can disrupt production and distribution, thereby affecting wholesale rice prices. This study aims to forecast wholesale rice prices in Indonesia using the Autoregressive Integrated Moving Average with Exogenous Variables (ARIMAX) model by incorporating ENSO indicators and hydrometeorological disaster variables. Monthly data from January 2015 to February 2026 were analyzed through stationarity testing, model identification, and parameter estimation, followed by model evaluation using Mean Absolute Percentage Error (MAPE). The results show that the best model is ARIMAX(1,1,0), which produces a MAPE value of 3.42%, indicating highly accurate forecasting performance. Despite this high accuracy, the influence of ENSO and disaster variables is relatively weak and occurs with a time lag, suggesting that rice prices are more strongly driven by internal market dynamics. These findings highlight that while climate and disaster factors contribute to price variations, their impact is limited in the short term.

Implementation of The Sequential Biclustering Method and Centroid-Based Imputation (Fuzzy C-Means) for Missing Values Imputation

Yolanda Azzahra — 2026-06-23

Missing values are a common issue in gene expression data and may significantly affect the accuracy of subsequent analyses if not properly handled. This study aims to implement and evaluate a missing value imputation method based on Sequential Biclustering and Centroid-Based Imputation with Fuzzy C-Means (FCM) on gene expression data of Type 2 Diabetes Mellitus patients. Missing values were generated under the Missing Completely At Random (MCAR) mechanism with missing rate ranging from 5% to 55%, and five replications were conducted for each missing rate. Biclustering was performed using the Cheng and Church algorithm to identify biclusters based on the Mean Squared Residue (MSR), followed by missing value estimation using FCM with 2 to 5 biclusters. The performance of the imputation method was evaluated using Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE) by comparing the imputed values with the original data. The results indicate that the configuration with 2 biclusters consistently produces the lowest and most stable MSE, RMSE, and MAE across all missing rates.

Analysis of the Influence of Macroeconomic Variables on the Movement of the Composite Stock Price Index using Nonlinear Autoregressive Distributed Lag

Michelle Amanda Godwinn — 2026-06-23

The stock market is one of the key indicators reflecting a country’s economic conditions. In Indonesia, the Composite Stock Price Index (CSPI) serves as a primary benchmark of domestic stock market performance. Movements in the CSPI are influenced by various macroeconomic variables, including real sector activity as represented by the Industrial Production Index (IPI) and the interest rate. Previous studies suggest that the relationship between the CSPI and macroeconomic variables is not always linear, but may exhibit asymmetric behavior in both the short run and the long run. In this context, asymmetry refers to differences in the effects of positive and negative changes in macroeconomic variables on the CSPI. To capture such nonlinear relationships, this study employs the Nonlinear Autoregressive Distributed Lag (NARDL) approach. The optimal lag selection based on the Akaike’s Information Criterion (AIC) identifies NARDL(12,12,11) as the preferred model specification. The estimation results indicate that the CSPI has a long-run relationship with both the IPI and the interest rate, as evidenced by the bound cointegration test. Furthermore, the relationship among these variables is asymmetric. Increases and decreases in the IPI have different effects on the CSPI, with strong evidence of long-run asymmetry, while short-run asymmetry remains present but weaker. Dynamically, declines in the IPI tend to trigger stronger and more volatile responses in the CSPI compared to increases in the IPI, whose effects generally materialize with a time lag. In contrast to the IPI, the interest rate does not exhibit long-run asymmetry. However, significant short-run asymmetry is observed. Decreases in the interest rate have a positive and significant effect on the CSPI after several periods, whereas increases in the interest rate do not show a significant short-run impact. Overall, decreases in industrial activity and decreases in the interest rate have a stronger effect on the CSPI than increases in these variables. This shows that the CSPI responds differently to positive and negative changes in both industrial activity and interest rates.

Path Analysis Model of the Influence of Mother’s Education and Employment on Stunting through Family Health Behavior as a Mediation Variable

Ade Irawan — 2026-06-23

Stunting is one of the health problems that is still a serious concern that has an impact on the growth and development and quality of life of children in the long term. WHO states that this condition is characterized by disturbances in the growth and development of children's physical, cognitive, and productivity. The condition of stunting in toddlers can be influenced by various factors, both direct and indirect factors. Maternal characteristics such as education and employment levels are important factors that can affect family health parenting patterns and behaviors which ultimately impact the nutritional status of children. This study aims to analyze the influence of maternal education and employment levels on the incidence of stunting in toddlers through family health behavior as a mediating variable. This study uses a quantitative approach with an observational analytical research design. The research sample was mothers who had toddlers who were selected using sampling techniques that were in accordance with the research criteria. Data collection was carried out through the distribution of questionnaires in a structured manner. Data analysis was carried out using path analysis to determine the direct and indirect influence between the variables studied. The results of the study show that maternal education and maternal work have a significant influence on family health behavior and the incidence of stunting in toddlers. The amount of direct influence of maternal education on stunting was 0.227, while the indirect influence mediated by family health behavior on stunting was 0.049. The magnitude of the direct influence of maternal work on stunting was 0.556, while the indirect influence mediated by family health behavior on stunting was 0.071. So that the total influence of maternal education and maternal work on the incidence of stunting was 0.376 and 0.627 respectively.

Chen-Burr XII Distribution

Laras Kirana Anindita — 2026-06-23

Survival analysis focuses on modeling the time until an event occurs, where the hazard function determines the flexibility of distributions for modeling lifetime data. The Chen distribution, defined by two positive parameters, can represent monotonically increasing and bathtub-shaped hazard functions. However, it cannot accommodate monotonically decreasing and unimodal (upside-down bathtub) hazard patterns, which are frequently encountered in failure risk and medical survival data. This study aims to construct a more flexible distribution capable of accommodating four hazard shapes: increasing, decreasing, unimodal, and bathtub-shaped using the Transformed-Transformer method. To address this limitation, the Burr Type XII distribution is used as the transformer distribution due to its ability to capture decreasing and unimodal hazard patterns, while the Chen distribution acts as the transformed distribution. The resulting Chen-Burr XII distribution has four positive parameters . Characteristics discussed in this study are the probability density function, cumulative distribution function, survival function, hazard function, and r-th moment. The maximum likelihood estimation approach is used to estimate parameter values. The Chen-Burr XII distribution is applied to survival data of gastric cancer patients' waiting time to death, exhibiting a unimodal hazard function. Model performance is evaluated using the Kolmogorov-Smirnov test and Akaike’s Information Criterion. The results show that the Chen-Burr XII distribution models data with a lower AIC value compared to the Chen distribution, offering a more flexible alternative for lifetime data with diverse hazard patterns.

Modeling the Percentage of Poor People in West Nusa Tenggara Using Truncated Spline Nonparametric Regression

Maryam Febrianti — 2026-06-23

Nonparametric regression approaches have received significant attention due to their high flexibility and ability to model data without relying on specific functional form assumptions. One widely applied method is the truncated spline, which effectively handles data behavior changes within specific sub-intervals. This study aims to model the Percentage of Poor People in West Nusa Tenggara Province from 2017 to 2025 using independent variables Average Years of Schooling (X1), Open Unemployment Rate (X2), Per Capita Expenditure (X3), and Life Expectancy (X4). The best model selection was performed by evaluating combinations of 1 to 3 knot points, using the minimum Generalized Cross Validation (GCV) as the primary criterion. Based on the analysis, the optimal truncated spline nonparametric regression model utilizes 1 knot point and degree 1, yielding a GCV value of 12.34126 and a coefficient of determination (R²) of 59.38%. The identified optimal knot points are 8.015 for X1; 3.07 for X2; 10310.5 for X3; and 68.095 for X4.