Jurnal Aplikasi Statistika & Komputasi Statistik

Separated Couples during the COVID-19 Outbreak: A Survival Support Vector Machine Analysis

2024-04-05T15:10:24+07:00

Introduction/Main Objectives: The separation between spouses has been rising noticeably in recent years in Palangka Raya, particularly during the COVID-19 outbreak. Background Problems: An analysis of time-to-event on those separations will be undertaken quantitatively using survival analysis by comparing the results yielded by Cox proportional hazards (PH) regression and non-parametric Survival Support Vector Machine (SUR-SVM). Novelty: This work suggests a feature selection method that looks for influencing elements related to the c-index by employing backward elimination. Research Methods: This study's data came from Indonesia's Supreme Court webpage, including a database of separation verdicts from the Palangka Raya Religious Court, spanning from April 2020 to March 2021. The response variables were the time-to-separation (marriage length until separation) (t) and the censored state of the occurrence (?). Finding/Results: Based on SUR-SVM, the factors contributing the most to the separation are the absence of children, unsteady employment of appellants, and finance motive as the primary reason. In terms of concordance index and Akaike Information Criterion (AIC), the SUR-SVM outperformed the Cox proportional hazard model. These values of SUR-SVM were 59.24 and 1899.78, respectively. SUR-SVM correctly classified 59.24% of separations based on the chronological order of events.

Early Study of LLM Implementation in Survey Interviews

2024-10-14T15:21:56+07:00

Introduction/Main Objectives: This research aims to conduct a preliminary study into the use of LLMs for extracting information to fill out questionnaires in survey interviews. Background Problems: BPS-Statistics Indonesia used paper-based questionnaires for interviews and is recently utilizing the Computer Assisted Personal Interviewing (CAPI) method. However, the CAPI method has some drawbacks. Enumerators must input data into the device, which can be burdensome and prone to errors. Novelty: This study uses a large language model (LLM) to extract information from survey interviews. Research Methods: This study utilizes a text-to-speech application to translate interview results into text. Translation accuracy is measured by the Word Error Rate (WER). Then the text was extracted using the ChatGPT 3.5 Turbo model. GPT-3.5 Turbo is part of the GPT family of algorithms developed by OpenAI. Finding/Results: The extraction results are formatted into a JSON file, which is intended to be used for automatic filling into the database and then evaluated using precision, recall, and F1-score. Based on research conducted by utilizing the Speech Recognition API by Google and the ChatGPT 3.5 Turbo model, an average WER of 10% was obtained in speech recognition and an average accuracy of 76.16% in automatic data extraction.

Quantile Regression with Constrained B-Splines for Modelling Average Years of Schooling and Household Expenditure

2024-09-24T14:46:02+07:00

Introduction/Main Objectives: Education serves as a driving force for the transformation of society to break the cycle of poverty. This study examines the relationship between average years of schooling and per capita household expenditure in Kalimantan Tengah Province in 2020. Background Problems: The method of estimating a regression model that is assumed to follow a certain form of regression equation such as linear, quadratic and others is called parametric regression. However, researchers often encounter difficulties in determining the model specification through data distribution, so the method used is nonparametric regression. Novelty: This research uses a quantile-based approach to explore how the impact of education on per capita expenditure varies across different levels of household education. This provides a more nuanced understanding of the relationship, showing not just whether education matters, but how its influence changes at different levels of educational attainment. Research Methods: The relationship between average years of schooling and per capita household expenditure is modeled using a quantile regression model with the constrained B-Splines method. Finding/Results: Based on the established classification, it can be concluded that an increase in the average years of schooling among household members tends to have a greater impact on raising per capita expenditure.

The Implementation of Geospatial Analysis on Hotel Occupancy Rate

2024-09-25T15:38:42+07:00

Introduction/Main Objectives: One of the main attributes of hotel selection and customer satisfaction is its location. Background Problems: Strategic location leads to higher demand for accommodation. Accommodation demand is reflected in hotel occupancy levels, which indicate the percentage of reserved rooms at a specific period. Novelty: This study aims to investigate the effect of spatial location on hotel occupancy rates by analyzing data collected in online hotel reservation applications. A study related to the effects of location and hotel occupancy has never been conducted in Indonesia. Research Methods: We use data from hotels located in the province of Yogyakarta, which contains 245 hotels spread over three regencies/cities, namely Yogyakarta City, Sleman Regency, and Bantul Regency. We conducted a spatial regression analysis, namely the Spatial Error Model (SEM), with a spatial weight matrix using a radius of 3.2 km. Finding/Results: We found that spatial locations affect the occupancy rates of hotels based on the online hotel reservation application that we observed. These spatial locations include the distance from the hotel to the airport, the distance from the hotel to the bus stop, and the number of nearby restaurants, offices, and hotels.

Comparison of Binary and Traditional Partial Least Squares Structural Equation Modeling: A Study on The Role of Multidimensional Poverty Dimension to Social Protection in Java Island

2024-10-14T15:16:14+07:00

Introduction/Main Objectives: The traditional Partial Least Squares Structural Equation Modeling (PLS-SEM) method uses an ordinary least squares regression approach that assumes that indicators must have a continuous scale. When the indicators are categorical, the use of traditional PLS-SEM becomes less appropriate. Background Problems: Multidimensional poverty consists of dimensions that are measured by a binary scale. The use of binary PLS-SEM is better than traditional PLS-SEM in modeling the effect of dimensions on social protection on Java Island. Novelty: The use of binary PLS-SEM with factor scores from the item response theory model applied to the role of dimensions of multidimensional poverty to social protection has not been carried out yet. Research Methods: This study introduces binary PLS-SEM, which is modified from traditional PLS-SEM by changing the data input using a tetrachoric correlation matrix. Finding/Results: Empirical results show that the binary PLS-SEM measurement model is better than traditional PLS-SEM. Evaluation of the structural model shows that the path coefficients of binary PLS-SEM are better than traditional PLS-SEM. Both approaches have an overall model fit. The order of multidimensional poverty dimensions that affect social protection are education, living standard, and health.

The Application of Partial Proportional Odds Model on Determinants Analysis of Household Food Insecurity Level in Papua, Indonesia

2024-09-26T14:55:04+07:00

Introduction/Main Objectives: Food insecurity in Papua, Indonesia, is still high. However, the study on that issue is limited. This research aims to analyze the determinants of food insecurity in Papua. Background Problems: An ordinal logistic regression can be used. However, this model generally requires the parallel lines assumption. However, somehow, the assumption is often violated. Novelty: This study used a model that relaxes the assumption of parallel lines. This model can capture the condition that some parameters are assumed to meet parallel lines and some do not. Research Methods: In this case, the partial proportional odds model was applied to find the determinant of household food insecurity status by using the National Socioeconomic Survey (SUSENAS) data. Finding/Results: The results show that a female head of household, age 60 years and above, junior high school education and below, has a higher tendency to be at least mildly food insecure, and the effect is the same for each level of food insecurity. Household heads who do not work, work in agriculture, and have household drinking water sources that are not feasible can aggravate the food insecurity level. Meanwhile, food assistance provided by the government influences reducing food insecurity levels.

Estimation of Gross Regional Domestic Product per Capita at the Sub-District Level in Bali, NTB, and NTT Provinces Using Machine Learning Approaches and Geospatial Data

2024-11-12T07:57:29+07:00

Introduction/Main Objectives: This study aims to estimate Gross Regional Domestic Product (GRDP) per capita at the sub-district level. Background Problems: Currently, GRDP per capita is calculated only at the district level by BPS. Novelty: This study estimates GRDP per capita at the sub-district level using a model developed at the district level, applying machine learning and linear regression methods. Research Methods: The model was constructed using geospatial data sourced from satellite imagery, OpenStreetMap, (Village Potential Statistics) PODES, directories of large mining companies, and directories of the manufacturing industry at the district level. Linear regression and machine learning methods, including neural networks, random forest regression, and support vector regression, were used to develop the model. The research focuses on three provinces: Bali, West Nusa Tenggara (NTB), and East Nusa Tenggara (NTT). Findings/Results: The best-performing model was support vector regression, with MAE and MAPE evaluations of 10.33 million and 26.11%, respectively. The results indicate that sub-districts with high GRDP per capita are typically urban areas that serve as economic hubs. The Williamson Index results show that districts in the eastern region have higher inequality levels compared to those in the western region.