Jurnal Statistika dan Komputasi

Pemodelan Indeks Kualitas Lingkungan Hidup di Indonesia dengan Spline Truncated dan MARS

2025-06-04T09:53:47+00:00

Background: Indonesia, endowed with abundant natural resources, faces substantial challenges in maintaining environmental quality amid rapid urbanization and economic growth. The 2022 Environmental Performance Index ranked Indonesia 164th out of 180 countries with a score of 28.2. Regionally, Indonesia ranked 22nd among 25 Asia-Pacific countries. The Environmental Quality Index (EQI), crucial for achieving the Sustainable Development Goals (SDGs), was recorded at 72.42 in 2022, classified as "fair." This condition underscores the need for in-depth analysis of key factors influencing environmental quality.

Objective: This study aims to examine significant factors affecting the Environmental Quality Index (EQI) across Indonesian provinces using appropriate nonparametric statistical methods.

Methods: A nonparametric regression approach, specifically the Multivariate Adaptive Regression Spline (MARS) and the truncated spline multipredictor model, was applied. Predictor variables included the Human Development Index (HDI), population density, access to proper sanitation, poverty rate, and Gross Regional Domestic Product (GRDP). Secondary data for 34 provinces in 2022 were sourced from the Central Bureau of Statistics and the Ministry of Environment.

Results: The truncated spline model demonstrated superior performance, achieving a minimal MSE of 5.63308, minimal GCV of 10.42, and R² of 82.63%, outperforming MARS, which yielded a minimal MSE of 7.685, GCV of 16.014, and R² of 79.3%. All predictor variables significantly influenced EQI.

Conclusion: Social and economic factors were found to significantly affect environmental quality. The truncated spline approach offers an effective modeling alternative, providing critical insights to support environmental policy development at the provincial level.

Peramalan Jumlah Barang Kereta Api di Indonesia Menggunakan Metode Seasonal Autoregressive Integrated Moving Average (SARIMA)

2025-06-04T09:55:08+00:00

Background: Freight transportation is an important part of the business run by PT Kereta Api Indonesia. To support effective strategic planning and infrastructure development, an accurate prediction of the amount of goods to be transported in the future is required. Therefore, historical data-based forecasting methods such as Seasonal Autoregressive Interated Moving Average (SARIMA) can be a relevant approach to predict the number of railway goods in Indonesia.

Objective: Obtain a suitable model to forecast the number of goods transported by rail transportation in Indonesia, and to determine the results of the forecasting.

Methods: This research uses the time series method with the Seasonal Autoregressive Integrated Moving Averang (SARIMA) model approach based on data characteristics that show seasonal patterns. SARIMA itself is able to integrate seasonal pattern components in the data and is able to effectively capture periodic and structural dynamics in seasonal data.

Results: The best model obtained is probabilistic SARIMA(0,1,1)(0,1,1)¹², using secondary data sourced from the Central Bureau of Statistics (BPS) in the range of January 2013 to March 2024. Forecasting for the next 12 months (April 2023 to March 2024) shows a Mean Absolute Percentage Error (MAPE) value of 8.03% which indicates that the level of forecasting accuracy is very good.

Conclusion: The probabilistic ARIMA(0,1,1)(0,1,1)¹² model can be used as a reliable reference in predicting the amount of goods transported through rail transportation in Indonesia.

Modeling the Satisfaction of Data Literacy Online Training for High School Teachers Using PLS SEM

2025-06-04T10:04:38+00:00

Background: The gap between before and after the pandemic is incredibly noticeable, especially in education. It mainly changes how schools operate their teaching and learning activities from offline to online. Indonesia must also implement online learning. The basic needs for data literacy in administration are strongly needed, such as inputting data for students' attendance, scores, and many more. Teachers need to improve their data literacy skills to help them evaluate and design new content structures for online teaching to meet students' required satisfaction. Therefore, the teachers’ training program in data literacy always needs to be updated.

Objective: This study aims to determine the factors that influence teacher satisfaction in data literacy online training programs

Methods: This study employs partial least square structural equation modelling (PLS-SEM) to analyze the factors influencing teacher satisfaction in online data literacy training programs for high school teachers.

Results: The results show that the instructor's guidance, support, module content, and experience positively influence learner satisfaction in online data literacy training. The PLS-SEM can explain 62.53% of learner satisfaction

Conclusion: Online training providers can consider these variables their primary focus when providing high-quality online training, especially in data literacy. The instructor guidance and support include instructor expertise, the assistance provided, and many more, and the module content and experience include a suitable syllabus for the learner and the ease of use of the learning system.

Clustering Job Seekers in Bojonegoro Using K-Means and Fuzzy K-Means

2025-06-04T10:05:55+00:00

Background: Job seekers are part of the labor force who are unemployed and actively looking for work. One of the efforts to address the rising number of job seekers is by expanding job openings or employment opportunities. Employment is an essential need for individuals to meet various aspects of life, ranging from basic needs to education and housing.

Objective: This paper aims to analyze the frequency distribution of job seeker attributes in Bojonegoro, compare the K-Means and Fuzzy K-Means methods in clustering sub-districts, determine the best clustering method, and describe frequency distribution for each formed cluster.

Methods: The methods used are K-Means and Fuzzy K-Means, both known for their ease of implementation and effectiveness in clustering large datasets by minimizing the average distance between data points within each cluster.

Results: The majority of job seekers in Bojonegoro in 2022 are aged 15–24, unmarried, and senior high school graduates, with males comprising 59.6% of the total. The clustering analysis, with an optimal k equal to 5, reveals five balanced groups with distinct variations in age, gender, and marital status, suggesting a range of employment needs among subgroups.

Conclusion: The findings indicate that most job seekers in Bojonegoro are young, male, unmarried, and secondary school graduates. The clustering process identified five relatively even groups, with K-Means slightly outperforming Fuzzy K-Means in cluster cohesion.

Clustering Analysis of School Student Distribution in Bojonegoro Regency with Kernel K-Means

2025-06-05T06:59:19+00:00

Background: Education is a fundamental aspect that plays an important role in determining the progress of a nation. In Indonesia, education equity remains a major challenge, particularly in relation to the gap between urban and rural areas. Bojonegoro District, with its diverse geographical and social characteristics, reflects this issue of uneven educational access.

Objective: This study aims to examine the 2022 distribution of study groups at different education levels (kindergarten, elementary school, junior high school, high school, and vocational school) in Bojonegoro using descriptive statistics, normality testing, and the Kernel K-Means clustering algorithm.

Methods: Data were tested for normality using the Kolmogorov-Smirnov method. The clustering was performed by comparing four kernel types (Dot, Polynomial, Gaussian, and Sigmoid) to determine the most effective approach based on the Average Within Cluster Distance (AWCD) and the elbow method principle.

Results: The findings indicate that the distribution of study groups at the kindergarten and elementary school levels is relatively even and follows a normal pattern. In contrast, the distributions at the junior high school, high school, and vocational school levels remain uneven. The Kernel K-Means algorithm with the Dot kernel produced the most optimal results, identifying five main clusters that reflect regional disparities in educational participation.

Conclusion: This study demonstrates the novelty of applying Kernel K-Means in the educational context to uncover spatial disparities. The resulting clusters offer valuable insights into education inequality in Bojonegoro. These insights can inform policymakers in designing more targeted, equitable, and data-driven education policies.