Jurnal Matematika, Statistika dan Komputasi

Comparison of Extreme Learning Machine (ELM) and Multi-Support Vector Machine (Multi-SVM) Methods in the Identification of Herbal Plants.

Luluk Sarifah, Lailiyatus Sa’adah, Iis Setiana

Hereditarily, Indonesian people have utilized herbal plants as ingredients for making traditional medicines, even with technological advances, they have been utilized in the pharmaceutical industry that are efficacious for health. There are around 2,039 species of herbal medicinal plants in Indonesia that sometimes have similarities, making it difficult to identify the types of herbal plants and switch to using more practical chemical drugs. The purpose of this study is to facilitate the identification of herbal plant types using machine learning methods and digital images by comparing the performance of the Extreme Learning Machine (ELM) and Multiclass Support Vector Machine (Multi-SVM) methods, so that the most effective and efficient method for identifying herbal plants can be obtained. The ELM method was created to overcome the weaknesses of feedforward artificial neural networks, especially in terms of learning speed, while the Multi-SVM method is a supervised machine learning algorithm that helps in classification problems and this method is also a further development of the SVM method. Based on the simulations that have been carried out, identification using the ELM method obtained accuracy for 5 types of herbal plant data, training and testing each with an accuracy of 100%. 10 types of herbal plant data, training data obtained 99% accuracy and testing data of 96.667%. 20 types of herbal plant data training data obtained 91% accuracy and testing data of 75%. 30 types of herbal plants training data obtained 80.33% accuracy and testing data of 68%. As for the Multi-SVM method in identifying 5 types of herbal plant data, training data accuracy was obtained. 52% and testing 66.667%. 10 types of herbal plants for training data obtained 39% accuracy and testing 30%. 20 types of herbal plants for training data obtained 23.5% accuracy and testing 23.33%. and 30 types of herbal plants for training data obtained 16.67% accuracy and testing 14.44%.

Optimasi Portofolio Berdasarkan Model Mean-Variance dengan Menggunakan Lagrange Multiplier pada Saham IDX30

Werry Febrianti, M. Naif Abdallah, Lutfi Mardianto

This study aims to optimize the portofolio of stocks included in the IDX30 index using the Mean-Variance model developed by Markowitz. The Lagrange Multiplier method is used in this study to determine the optimal fund allocation by minimizing risk and optimizing expected return. The data used is the daily closing price of stocks from 15 companies listed in the IDX30 index over the last five years (2019-2024). The results show that the Mean-Variance and Lagrange Multiplier methods are effective in identifying the optimal portofolio that can minimize investment risk while maximizing returns.

Nonparametric Regression with A Fourier Series Approach on Food Security Index in East Java

Nur Afriani, Sifriyani, Meirinda Fauziyah

Nonparametric regression is a regression model approach used when the regression curve is
unknown. One of the estimators that can be used to estimate nonparametric regression
approaches is the Fourier series. The Fourier series is a trigonometric polynomial that has
flexibility so that it can adjust effectively to the local properties of the data. The purpose of the
study is to estimate a nonparametric regression model with a Fourier series approach and
obtain factors that affect the food security index in East Java in 2022. Factors thought to be
influential are the prevalence of stunted toddlers, rice production, the percentage of
households with access to clean water, the percentage of poor people, and life expectancy.
Based on the results of the study, it was obtained that the best model of nonparametric
regression of the Fourier series is a model with 3 oscillations which is shown with a minimum
Generalized Cross-Validation (GCV) value is 7.60 and a determination coefficient value
is 90.89%.

Implementation Of Mathematical Logic Based Artificial Intelligence In Breast Cancer Diagnosis: A Literature Review

Andi Sitti Nur Fatimah Madaeng

Abstract

Introduction: Breast cancer is the most common type of cancer in Indonesia so accurate diagnosis is required. Implementation of mathematical logic based on Artificial Intelligence (AI) can be a solution for more accurate diagnosis. Objective: To find out the AI-based mathematical logic approach in breast cancer diagnosis. Methods: The literature study was conducted using search tools from several databases, namely ScienceDirect, PubMed, Google Scholar with keywords within the study limit of the last 10 years in English. There were 9 literatures reviewed in this literature review. Discussion: AI through Convolutional Neural Networks (CNNs), Naïve Bayes (NB), and Support Vector Machine (SVM) systems. have diagnostic capabilities. CNN DLs has an accuracy value of 0.9, specificity of 0.96, and sensitivity of 0.86 whereas, SVM has an accuracy of 98% and NB has an accuracy value of 97.82%. Conclusion: AI in breast cancer diagnostics improves accuracy, speed, efficiency of diagnosis, and makes treatment more personalized.

Simulation of Distance Metrics on K-Means Clustering for Strategic Market Segmentation Private Universities

Regita Permata

K-Means clustering is a widely used method for grouping data based on feature similarities. The choice of distance metric in K-Means significantly impacts the quality and efficiency of clustering results. This study evaluates the performance of various distance metrics—Euclidean, Cityblock (Manhattan), Canberra, and Mahalanobis—in the context of clustering for university market segmentation. The analysis utilizes simulation data and university segmentation strategy factors, including quantity (accounts, registration, graduation, PINs) and quality (social, science, scholastic values) components, as well as income and distance. Results show that Euclidean and Cityblock metrics yield well-defined clusters with good separation and low execution times, making them suitable for most tasks. Conversely, Mahalanobis distance, despite its consideration of data correlations, is less practical due to higher computational costs. The Canberra distance, while producing the most compact clusters, does not enhance overall clustering quality. For strategic market segmentation in private universities, Clusters 0 and 1 emerge as the most promising targets. Cluster 0, characterized by high-income and distant students, is ideal for premium programs, while Cluster 1, with diverse student profiles, offers broader marketing opportunities. Ultimately, the selection of an appropriate distance metric should align with specific task requirements, balancing clustering quality with computational efficiency.

Analysis of Public Interest in Digital Payment Using Google Trends

Aqilla Haya

This study examines the adoption of digital payments in Indonesia after the pandemic using Google Trends data. We found that the search trend for “digital payment” has been fluctuating, with the average peak for each keyword being at the end of 2023. The practicality of using digital payment and the presence of various promos, especially for digital payment users, has increased public interest in making cashless payments. However, the ease of access of financial and banking companies is expected to improve security in digital payment applications. With SWOT analysis, we found that the presence of various promos has increased public interest in making cashless payments. This payment method also provides opportunities to digitize and automate to improve service efficiency and effectiveness. On the other side, this payment method still often leads to fraud on behalf of the application and hacking of user data through various techniques such as phishing, scams, and many more.

The Statistical Perspective of Dengue Hemorrhagic Fever (DHF) in West Java: Insights from Two-Way Random Effects Model

Ghiffari Ahnaf Danarwindu

The Indonesian Ministry of Health has reported an alarming increase in Dengue Hemorrhagic Fever (DHF) cases, particularly on the island of Java. In West Java province alone, there were 39,623 confirmed cases, making it the highest incidence in the country. Given this trend, collaborative research and surveillance efforts are crucial to understanding and managing DHF cases in Indonesia. From a modeling perspective in statistics, especially in panel data regression models, dengue fever is still commonly discussed. The panel data regression model in dengue fever cases will provide new insights in modeling and the field of panel data in the dengue fever data. This research goals to identify the most appropriate random effects model for estimating a dataset with four different variables. The research seeking panel data modeling on the influence of population density, percentage of poor people, percentage of adequate drinking water, and adequate sanitation rates on DHF cases in West Java province. This method emphasizes selecting the best model from one-way and two-way Random Effects Models (REM) and then identifying what factors influence the increase of DHF cases in West Java province. The best model obtained is a two-way Random Effects Model with three significant variables. Based on the selected variables in the model, it can be said that West Java Province needs to pay attention to the distribution of housing in each district and economic activity in each district because population density is an important concern studied by the local government.

Implementation of the Sarimal Statistical Algorithm for Sales Prediction in a Nigerian Medium Scale Enterprise

Olumuyiwa James Peter

A sales prediction is a projection of future sales figures for a business over a given period of time. Forecasts are predicated on conjectures drawn from examining the trends or behavior of historical sales data. Sales forecasting has an impact on most of the factors that lead to impact optimization, profit maximization, and cost minimization. An efficient forecasting system can help a business raise profits, increase equipment utilization, reduce inventory, and achieve more changeability. But the majority of medium-sized businesses, like the Nigerian fast-food sector, have not yet adopted this technology or business model, which results in poor planning and ultimately, early closure or bankruptcy. This led to the research's objective, which is to apply the statistical method to anticipate sales in a medium-sized fast-food company. Data collection, data exploration, and statistical model creation were used to accomplish this. The machine learning model was constructed using the statistical approach known as Seasonal Autoregressive Integrated Moving Average, or SARIMA. To use its features, the model was made available as a web application. The outcome demonstrates that, if implemented, it would allow Nigeria's fast-food industry, as well as any other medium-sized industry in the nation, to make well-informed projections, maximize profits and as a result guaranteeing their survival in the competitive business environment of the developing world.

Simulation Modeling Incomplete Treatment Impact on Tuberculosis Transmission

Muna Afdi Muniroh

Among the most common diseases globally is tuberculosis (TB). The spread dynamics of TB are formulated in the form of a mathematical model with five subpopulation densities, namely, susceptible individuals, latent individuals, TB active individuals, treated individuals, and recovered individuals. The existence of an equilibrium point has a basic reproduction number boundary condition. R0 is a key metric for understanding the potential for disease transmission and is obtained from the next generation matrix. Stability analysis for TB models is investigated by determining the criteria for the local stability of equilibrium points. After that, a sensitivity analysis is conducted to identify TB model parameters that most affect R0 value. The solution behavior of the TB model is shown by graphs generated numerically with the Runge-Kutta fourth-order method and Matlab software.

Spatial Extreme Value Modeling of Seawater Wave Height in Banyuwangi with Smith and Geometric Gaussian Models

Annisa Fathimah Azzahra, Ulil Azmi, Prilyandari Dina Saputri

Banyuwangi faces serious challenges related to Coastal Disasters, with a high risk of extreme waves and abrasion by 2022. Therefore, an understanding of wave characteristics and behavior is essential for disaster mitigation. Spatial analysis was conducted in four coastal disaster-prone locations in Banyuwangi Sea using the Spatial Extreme Value method with the Max-Stable Process approach of Smith and Geometric Gaussian models. The data analyzed is seawater wave height data for the period January 1, 2019 - December 31, 2023. The Block Maxima method was used to select extreme values, with 80% of the data allocated for training data and 20% for testing data. The training data follows a stationary Generalized Extreme Value distribution, with no trend. Then, the training data is transformed to frechet margin units and the extremal coefficients are calculated. The resulting extremal coefficients indicate a fairly strong dependency between locations. The best trend surface model is a model with location parameters that have a latitude coordinate factor and scale that has a longitude coordinate factor. The estimation of the spatial parameters of the Smith and Geometric Gaussian models for each location was carried out using the Nelder-Mead and BFGS Quasi Newton numerical iteration methods, and the frechet and GEV margin unit level returns were calculated. Model evaluation is done by calculating the MAPE, SMAPE, and RMSE values for each model and iteration method. The MAPE, SMAPE, and RMSE values of the Smith model with the Nelder-Mead iteration method are 0.23912%, 0.23873%, and 0.01575, while the Geometric Gaussian model is 0.32274%, 0.32212%, and 0.02001. The estimated return level of the Sea Wave Height in the next five years at the four water locations is classified in the extreme category.