cover
Article Per Year (5 Year)
All
Home Page
OAI Link
Editorial Team
Contact
Reviewer
Google Scholar
Contact Name
Muh. Isbar Pratama
Contact Email
isbarpratama@unm.ac.id
Phone
+6285399692435
Journal Mail Official
jmathcos@unm.ac.id
Editorial Address
Kampus Parangtambung UNM, Jl. Dg. Tata Raya Prodi Matematika Lt. 3 Gd FG Jurusan Matematika FMIPA
Location
Kota makassar,
Sulawesi selatan
INDONESIA
Journal of Mathematics, Computation and Statistics (JMATHCOS)
Published by Universitas Negeri Makassar
ISSN : 24769487     EISSN : 27210863     DOI : https://doi.org/10.35580/jmathcos
Core Subject : Education,
Mathematics
Fokus yang didasarkan tidak hanya untuk penelitian dan juga teori-teori pengetahuan yang tidak menerbitkan plagiarism. Ruang lingkup jurnal ini adalah teori matematika, matematika terapan, program perhitungan, perhitungan matematika, statistik, dan statistik matematika.
Arjuna Subject : Matematika - Matematika Terapan
Articles 241 Documents
 21   22   23   24   25 
A Systematic Simulation Study of Semiparametric Spline Estimators for Nonlinear Data Structures in R Rahmat Hidayat; Aswi Aswi; Zakiyah Mar'ah
Journal of Mathematics, Computations and Statistics Vol. 9 No. 2 (2026): Volume 09 Issue 02 (June 2026)
Publisher : Jurusan Matematika FMIPA UNM

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35580/7k3vqe29

Abstract

In many real-world applications, the assumption of linearity in classical regression models is often violated, leading to model misspecification and inaccurate estimation when data exhibit complex nonlinear patterns. Although nonparametric approaches provide flexibility, they frequently suffer from poor interpretability and instability in high-dimensional settings. To address these limitations, this study examines the implementation of semiparametric spline regression as a flexible yet interpretable alternative. The model integrates a linear component for certain predictors and a spline-based nonparametric component to capture local data fluctuations. Through a simulation study using the R programming language, the performance of the spline estimator was evaluated based on the Generalized Cross Validation (GCV) criterion for optimal knot selection. The results demonstrate that the semiparametric spline model achieves superior accuracy, with a coefficient of determination (R²) reaching 97.35%, compared to 81.18% for the classical linear model. In addition, the Mean Square Error (MSE) is significantly reduced from 2.158 to 0.303. Residual diagnostic analysis confirms that the model satisfies normality and homoscedasticity assumptions. These findings highlight the effectiveness of spline-based semiparametric regression in modeling complex nonlinear data structures.
Advancing Panel Data Analysis: A Dual-Evidence Assessment of Linear Mixed Models Rahayu, Melania Dwi; Djuraidah, Anik; Kurnia, Anang
Journal of Mathematics, Computations and Statistics Vol. 9 No. 2 (2026): Volume 09 Issue 02 (June 2026)
Publisher : Jurusan Matematika FMIPA UNM

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35580/8h6xxf09

Abstract

This study evaluates and compares seven panel data model specifications in capturing temporal and cross-sectional variation using both simulated and empirical data. Panel data is employed for its ability to simultaneously account for heterogeneity across units and temporal dependence over time. In the first stage, Monte Carlo simulations assess model performance under controlled temporal structures, including AR(1), AR(2) and MA(3) processes. In the second stage, the models are applied empirically to poverty data across regencies and cities in East Java from 2012 to 2022. Simulation results are indicate that models explicitly incorporating stochastic temporal dynamics achieve the lowest RMSE, while specifications treating time merely as a covariate consistently underperform. Empirical results show that two-way fixed effects models controlling for persistent unit heterogeneity and common year effects provide the best predictive performance. Overall, findings highlight that appropriately modelling temporal variation is crucial for accurate panel data predictions, and the comparative evaluation offers guidance for selecting suitable model specifications in applied settings.
Forecasting Sharia Stock Prices Using Hybrid STL Decomposition–LSTM and SARIMA Models: A Case Study of PT Semen Indonesia (Persero) Tbk Indrawan Indrawan; Muhammad Azka; Putri Amalia
Journal of Mathematics, Computations and Statistics Vol. 9 No. 2 (2026): Volume 09 Issue 02 (June 2026)
Publisher : Jurusan Matematika FMIPA UNM

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35580/2ane1252

Abstract

Stock price forecasting is a challenging task due to the complex, nonlinear, and dynamic nature of financial time series data. This study aims to develop a hybrid forecasting model by integrating Seasonal and Trend Decomposition using Loess (STL) with Long Short-Term Memory (LSTM) and to compare its performance with the Seasonal Autoregressive Integrated Moving Average (SARIMA) model as a linear benchmark. The empirical analysis is conducted using daily closing price data of PT Semen Indonesia (Persero) Tbk (SMGR) over the period from March 2018 to March 2026. The proposed approach applies STL decomposition to separate the time series into trend, seasonal, and residual components, enabling the LSTM model to capture nonlinear patterns more effectively. Forecasting performance is evaluated using the Mean Absolute Scaled Error (MASE) on an out-of-sample testing dataset. The results show that the hybrid STL–LSTM model achieves superior accuracy, with a MASE value of 0.4738, significantly outperforming the SARIMA model, which yields a MASE value of 2.7073. In contrast, the SARIMA model produces overly smooth forecasts and fails to capture short-term fluctuations and nonlinear dynamics present in the data. These findings indicate that the integration of STL decomposition and LSTM provides a more effective and flexible framework for modeling complex financial time series. The proposed model not only improves forecasting accuracy but also produces stable and reliable predictions, making it suitable for practical applications in stock price forecasting.
Forecasting Acute Respiratory Infection Incidence in South Sulawesi Province Through a Hybrid ARIMA–RBFNN Model Muthia Ramadhani Rafli; Muhammad Abdy; Wahidah Sanusi
Journal of Mathematics, Computations and Statistics Vol. 9 No. 2 (2026): Volume 09 Issue 02 (June 2026)
Publisher : Jurusan Matematika FMIPA UNM

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35580/ca608g80

Abstract

Abstract. Among all notifiable diseases in Indonesia, Acute Respiratory Infection (ARI) consistently registers the highest national burden of illness. Within South Sulawesi Province alone, the eight-month tally from January through August 2023 surpassed 320,942 confirmed cases, underscoring the critical need for reliable case-number projections to guide evidence-based health-service planning. The present work constructs a time series forecasting framework that integrates ARIMA (Autoregressive Integrated Moving Average) with a Radial Basis Function Neural Network (RBFNN) under the hybrid paradigm proposed by Zhang (2003). Monthly ARI incidence data spanning January 2014 to December 2024 provided 132 observations in total. Following a chronological split, the first 96 data points (January 2014–December 2021) served as the training set and the remaining 36 (January 2022–December 2024) as the hold-out evaluation set. ARIMA captured the linear dynamics of the series, whereas RBFNN was subsequently applied to the ARIMA residuals to account for any nonlinear structure that remained unexplained. Minimum-AIC model selection identified ARIMA(2,1,2) as the most suitable linear specification. For the RBFNN stage, a four-lag input vector—derived from the partial autocorrelation function—combined with four hidden units and a multiquadratic basis function delivered the best generalisation performance. Assessed against MAPE, RMSE, and R², the standalone ARIMA(2,1,2) attained 14.19%, 5038.37, and 0.6275, respectively; RBFNN alone produced 15.47%, 4714.93, and 0.5479; and the Hybrid ARIMA–RBFNN yielded 16.11%, 5014.99, and 0.6309. The superior R² of the combined model demonstrates its enhanced capacity to account for data variability. Because all three models returned MAPE values below the 20% threshold, they qualify as good predictors under the Lewis (1982) classification scheme. On this basis, the hybrid approach is put forward as the preferred tool for ARI early-warning and surveillance operations in South Sulawesi.
Clustering Indonesian Provinces Based on TKA Scores Using PCA-Based K-Means Clustering with Biplot Visualization Nurul Qisthi; Fajriatus Sholihah
Journal of Mathematics, Computations and Statistics Vol. 9 No. 2 (2026): Volume 09 Issue 02 (June 2026)
Publisher : Jurusan Matematika FMIPA UNM

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35580/dkptsq64

Abstract

The Academic Ability Test (TKA) serves as a tool for assessing students' competency levels and facilitates the comparison of academic performance among various provinces in Indonesia. The data from TKA, characterized by their multidimensional nature and intercorrelations, necessitate the use of a multivariate analytical strategy. This study aims to cluster the provinces of Indonesia based on TKA scores by using K-Means clustering in conjunction with Principal Component Analysis (PCA) and biplot visualization. The dataset comprises TKA scores collected from 16 subjects spanning 38 provinces. PCA is utilized to convert the variables into orthogonal principal components, effectively minimizing the influence of inter-variable correlations. Following this, clustering is executed through K-Means, where the ideal number of clusters is established by analyzing the pseudo-F statistic. The results indicate that the provinces can be divided into two distinct clusters: one characterized by relatively low academic achievement and the other by high academic achievement. The biplot visualization indicates that the differentiation of clusters is mainly driven by the first principal component, reflecting overall scholarly achievement. In conclusion, substantial disparities in academic achievement across Indonesian provinces are identified. The integration of PCA and clustering produces a more robust and interpretable grouping structure by accounting for inter-variable correlations. These findings provide a data-driven basis for designing targeted educational policies to enhance equity and improve overall education quality.
Comparison of Bayesian Spatial Leroux CAR Models with Poisson and Binomial Likelihoods for Modeling Stunting Cases Rika Saniarti; Aswi Aswi
Journal of Mathematics, Computations and Statistics Vol. 9 No. 2 (2026): Volume 09 Issue 02 (June 2026)
Publisher : Jurusan Matematika FMIPA UNM

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35580/nykfnx57

Abstract

Stunting remains a chronic nutritional problem requiring precise spatial mapping to support effective policy interventions in East Java, Indonesia. Spatial disease mapping commonly applies the Poisson distribution with Conditional Autoregressive (CAR) effects; however, the Poisson distribution is sensitive to overdispersion. Alternatively, the Bayesian Spatial CAR model with a Binomial likelihood may offer a better framework, yet empirical comparisons remain limited. This study compares the performance of Bayesian Spatial Leroux CAR models with Poisson and Binomial likelihoods in modeling stunting cases and identifies associated factors. The data include stunting cases across 38 districts in East Java (2024) and predictors: low birth weight (LBW), prematurity, exclusive breastfeeding, complete basic immunization (CBI), pneumonia, and diarrhea. Performance was evaluated using the Deviance Information Criterion (DIC) and the Watanabe–Akaike Information Criterion (WAIC). Results indicate significant spatial dependence. The Bayesian Spatial Leroux CAR model with a binomial likelihood outperforms the Poisson-based model. LBW, exclusive breastfeeding, CBI, pneumonia, and diarrhea are significantly associated with stunting. Kediri Regency exhibits the highest relative risk (RR), followed by Probolinggo Regency and Batu City, while Kediri City and Ponorogo Regency show the lowest RR
On the Mixture Non-Homogeneous Poisson Process to Model Spatial Heterogeneity of Earthquake Occurrences Nurhalisa Nurhalisa; Nur Iriawan; Achmad Choiruddin
Journal of Mathematics, Computations and Statistics Vol. 9 No. 2 (2026): Volume 09 Issue 02 (June 2026)
Publisher : Jurusan Matematika FMIPA UNM

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35580/5n6g9p44

Abstract

Earthquake occurrences exhibit complex spatial patterns influenced by geological structures such as faults, subduction zones, volcanic activity, and soil characteristics. Conventional approaches often assume homogeneous intensity, which may fail to capture spatial variability and underlying heterogeneity in seismic processes. Therefore, this study aims to model and compare earthquake occurrences in the Nusa Tenggara region during 2010–2025 using the Homogeneous Poisson Process (HPP), Non-Homogeneous Poisson Process (NHPP), and a Mixture Model-based Poisson process to identify the most appropriate modelling framework. The analysis is conducted using a spatial grid approach (10×15 cells), incorporating geophysical covariates including distances to faults, subduction zones, volcanoes, and soil conditions. Parameter estimation is performed within a Bayesian framework using Markov Chain Monte Carlo (MCMC) methods with consistent settings across all models. The results show that the NHPP model captures spatial variability better than HPP, with subduction distance, volcanic activity, and soil characteristics identified as significant factors, while fault distance is not statistically significant. However, the mixture model provides substantially improved model fit, revealing the presence of two latent components that represent different seismic patterns, with estimated proportions of 84.6% and 15.4%, respectively. Based on model comparison using the Widely Applicable Information Criterion (WAIC), the mixture model yields the lowest value (866.05), indicating superior predictive performance. In conclusion, incorporating both spatial non-homogeneity and latent heterogeneity leads to a more flexible and accurate representation of earthquake occurrences. It is recommended that future studies consider more advanced spatial or hierarchical modelling approaches to further enhance predictive accuracy
Rice Price Forecasting in South Sulawesi Using Neural Network Autoregression (NNAR) Hardianti Hafid; Isma Muthahharah
Journal of Mathematics, Computations and Statistics Vol. 9 No. 2 (2026): Volume 09 Issue 02 (June 2026)
Publisher : Jurusan Matematika FMIPA UNM

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35580/vpzsvb04

Abstract

Rice is a strategic food commodity in Indonesia due to its role as the main staple food and its significant contribution to inflation and economic stability. Fluctuations in rice prices directly affect purchasing power and are often used as an important indicator in assessing macroeconomic conditions. At the regional level, South Sulawesi plays a crucial role as one of the national food barns, where price dynamics may influence food availability and distribution, particularly in Eastern Indonesia. However, rice price data often exhibit non-linear patterns and sudden fluctuations, making accurate forecasting a challenging task. This study aims to evaluate the performance of the Neural Network Autoregression (NNAR) model in forecasting monthly rice prices in South Sulawesi. The study uses secondary time series data consisting of 61 observations from January 2021 to January 2026. The NNAR model is applied to capture non-linear relationships using lag-based inputs within a feed-forward neural network framework. The model performance is evaluated using Mean Absolute Percentage Error (MAPE) under several data splitting scenarios. The results show that the best model is NNAR (1,3) with a data split of 80% training and 20% testing, producing a MAPE value of 3.572%, which indicates excellent forecasting ability. The forecasting results suggest that rice prices are expected to remain relatively stable with a slight downward trend in the upcoming period. Overall, the NNAR model demonstrates strong capability in capturing the underlying patterns of rice price data and provides reliable forecasting performance. This study contributes to the development of time series forecasting methods and provides useful insights for policymakers in managing food price stability.
Semiparametric Spline Regression with Moving Average Smoothing Under Heteroscedastic Errors for Childhood Stunting Prevalence in Indonesia Jiran Julita; Jerry Dwi Trijoyo Purnomo; I Nyoman Budiantara
Journal of Mathematics, Computations and Statistics Vol. 9 No. 2 (2026): Volume 09 Issue 02 (June 2026)
Publisher : Jurusan Matematika FMIPA UNM

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35580/cgmfwz19

Abstract

Stunting prevalence remains a critical public health issue in Indonesia due to its long-term impact on child development and human capital. This study aims to model stunting prevalence using a semiparametric regression approach with truncated spline functions, which allows for capturing both linear and nonlinear relationships between variables. The percentage of poor population is treated as the parametric component, while immunization coverage is modeled as the nonparametric component. Model estimation is initially performed using Ordinary Least Squares, where the best model is obtained with two knot points and a minimum Generalized Cross Validation value of 21.74771. However, residual diagnostics indicate the presence of heteroscedasticity. To address this issue, Weighted Least Squares with a moving average approach is applied. The results show that the optimal weighted model uses three knot points with a cubic spline, producing a lower GCV value of 14.48820 and a higher coefficient of determination of 86.78 percent compared to 61.98 percent in the OLS model. Furthermore, all residual assumptions are satisfied under the weighted approach. These findings indicate that the WLS method with moving average provides a more accurate, stable, and reliable model for analyzing stunting prevalence.  
Comparison of Support Vector Regression and Random Forest Methods for Rainfall Prediction in Makassar City Ilmadinah Kadir; Wahidah Sanusi; Ja'faruddin Ja'faruddin
Journal of Mathematics, Computations and Statistics Vol. 9 No. 2 (2026): Volume 09 Issue 02 (June 2026)
Publisher : Jurusan Matematika FMIPA UNM

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35580/e1mjfd50

Abstract

This study aims to compare the performance of Support Vector Regression (SVR) and Random Forest (RF) methods in predicting daily rainfall in Makassar City and to identify the most influential meteorological factors. The dataset consists of daily climate data from 2019 to 2024, including rainfall as the response variable and temperature, humidity, wind speed, and sunshine duration as predictor variables. Data preprocessing was conducted through missing value imputation, time-series structuring, and normalization using the Z-score method for the SVR model. The SVR model was developed using several kernel functions, including linear, polynomial, radial basis function (RBF), and sigmoid, with hyperparameter tuning performed using grid search and k-fold cross-validation. Meanwhile, the Random Forest model was constructed using bootstrap aggregation and random feature selection, with optimal parameters determined based on the minimum out-of-bag (OOB) error. The results show that the SVR model with the RBF kernel achieved the best performance, with RMSE of 16.52 mm and MAE of 9.01 mm, outperforming the Random Forest model, which produced RMSE of 18.15 mm and MAE of 10.93 mm. Furthermore, feature importance analysis indicates that humidity and temperature are the most dominant variables influencing rainfall. Therefore, the SVR method is more accurate and reliable for rainfall prediction in Makassar City.

Page 23 of 25 | Total Record : 241

 21   22   23   24   25 

Filter by Year

2018 2026


Reset
Filter By Issues
All Issue Vol. 9 No. 2 (2026): Volume 09 Issue 02 (June 2026) Vol. 9 No. 1 (2026): Volume 09 Issue 01 (March 2026) Vol. 8 No. 2 (2025): Volume 08 Nomor 02 (Oktober 2025) Vol. 8 No. 1 (2025): Volume 08 Nomor 01 (April 2025) Vol. 7 No. 2 (2024): Volume 07 Nomor 02 (Oktober 2024) Vol. 7 No. 1 (2024): Volume 07 Nomor 01 (April 2024) Vol. 6 No. 2 (2023): Volume 06 Nomor 02 (Oktober 2023) Vol. 6 No. 1 (2023): Volume 06 Nomor 01 (April 2023) Vol. 5 No. 2 (2022): Volume 05 Nomor 02 (Oktober 2022) Vol. 5 No. 1 (2022): Volume 05 Nomor 01 (April 2022) Vol. 4 No. 2 (2021): Volume 04 Nomor 02 (Oktober 2021) Vol. 4 No. 1 (2021): Volume 04 Nomor 01 (April 2021) Vol. 3 No. 2 (2020): Volume 03 Nomor 02 (Oktober 2020) Vol. 3 No. 1 (2020): Volume 03 Nomor 01 (April 2020) Vol. 2 No. 02 (2019): Volume 02 Nomor 02 (Oktober 2019) Vol. 2 No. 01 (2019): Volume 02 Nomor 01 (April 2019) Vol. 1 No. 02 (2018): Volume 01 Nomor 02 (Oktober 2018) Vol. 1 No. 01 (2018): Volume 01 Nomor 01 (April 2018) More Issue