Jurnal Matematika, Statistika dan Komputasi

Classification of Unisba Students' Graduation Time using Support Vector Machine Optimized with Grid Search Algorithm

Ilham Faishal Mahdy

Support Vector Machine is a classification method that finds the optimal hyperplane to separate two data classes. SVM has much better generalization performance than other methods. However, SVM needs to improve in determining hyperparameter values. Therefore, parameter optimization is necessary to determine the optimal hyperparameter value. Grid search is one of the parameter optimization methods that can improve the quality of SVM models. This study aims to assess the level of accuracy in predicting student graduation times by using five features that affect it. Through this study, it showS that the resulting SVM model is quite accurate. By utilizing the results of SVM modelling, UNISBA is expected to improve the quality of graduates. The risk of delays in graduation can be considered early by paying attention to the background and achievements of students.

A STOCK PRIZE FORECASTING OF BAYAN RESOURCES TBK USING LONG SHORT TERM MEMORY DENGAN VARIABEL EKSOGEN

charles fernando

Stock are one of the financial instruments with the largest number of investors. Stocks have a large rate of return but are accompanied by large risks due to various exogenous variables. This causes the need for forecasting regarding stock prices so that investors know when it is time to invest and get maximum profits. This forecasting can be done using the Long Short Term Memory (LSTM) method because it has an architecture that can retain or discard information that is not needed in long-term historical data. In this study, LSTM is used to forecast the daily closing price of PT Byan Resources Tbk from 29 December 2020 to 29 December 2023 with 3 conditions, namely using exogenous variables in the form of open, low, volume, and exchange rate, exogenous variables that only use open and low, and do not use exogenous variables. The results of the study showed that the best prediction model was obtained using 2 exogenous variables with the best prediction model were 250 epochs and 32 batch sizes with a MAPE value of 1.27395% and a MAPE for forecasting of 4.44%.

Keywords: stock, Forecasting, LSTM, exogenous variable

GAMMA REGRESSION MODELING

Mariatul Qiftiyah Wulandari

The gamma regression model is a nonlinear regression model developed from generalized linear models (GLMs) and can be used to model the relationship between one or more predictor variables and response variables that have positive right-skewed continuous data with a gamma distribution. This study aims to get the best gamma regression model to model the percentage of poor people in districts/municipalities on Kalimantan Island in 2022 and get factors that have a significant effect on the percentage of poor people in districts or cities on Kalimantan Island based on the best gamma regression model. Parameter estimation in the gamma regression model uses the maximum likelihood estimation method and Fisher scoring. Parameter significance testing includes simultaneous and partial tests. The simultaneous test uses the likelihood ratio test, while the partial test uses the Wald test. The results showed that the best gamma regression model obtained by the backward elimination method was a gamma regression model with three predictor variables. The factors that have a significant effect on the percentage of poor people in districts/municipalities in Kalimantan Island based on the best gamma regression model are life expectancy, average years of schooling, and the percentage of households that have access to decent drinking water sources.

COVID-19 VACCINATION IMPACT ON ASEAN STOCK WITH SPATIAL DEPENDENCY: A COMPARISON OF PANEL AND GEOGRAPHICALLY WEIGHTED REGRESSION

Marizsa Herlina

Research about various policies and responses toward COVID-19 cases and its impact on stocks has grown recently. It shows that spatial influence is one of the keys in this research. The pandemic is not free from spatial dependence regarding how it indirectly impacts a country’s economy. Each country has different policies to handle COVID-19, such as lockdowns and vaccination. WHO stated that all countries require vaccination to build human immunity against COVID-19 in the future. Naturally, ASEAN implemented this policy; thus, it is crucial to see the extent of the impact of vaccination on the ASEAN economy. However, the residuals have heterogeneity problems when using the panel regression model. One of the reasons is that there is spatial dependence, especially when modeling the COVID-19 pandemic. Therefore, comparing panel regression with a geographically weighted regression panel (GWR-Panel) is substantial when exploring the reaction of stock returns to vaccination and positive cases of COVID-19 in Indonesia, Malaysia, Singapore, and Thailand.

Comparison of Classification Methods with Resampling Techniques on Chronic Kidney Disease Incidence in North Kalimantan

Putri Ramatillah Ramadhana

Chronic kidney disease is a non-communicable disease that has become a global issue. In Indonesia, North Kalimantan Province has the highest prevalence of chronic kidney disease, reaching 0.64%. The productive-age population in this province is one of the groups vulnerable to this disease. Although chronic kidney disease occurs more frequently in women, its progression is faster in men, with a prevalence of 18.41% of men undergoing hemodialysis compared to 0% of women. This is upsetting because more men are employed and have higher productivity. Therefore, this research aims to identify important variables contributing to chronic kidney disease in productive-age men through a classification modelling and the application of the best resampling technique. Resampling techniques are necessary due to the imbalance in the dataset. The resampling techniques compared are SMOTE and SMOTE + Tomek Links, with Random Forest, Naïve Bayes, logistic regression, and Firth logistic regression classification methods. The research findings indicate that the best classification method is Naïve Bayes with the best resampling technique being SMOTE. Important variables influencing chronic kidney disease in productive-age men include age, energy drink consumption, regional classification, instant food consumption, and sweet beverage consumption.

Determinants of Tax Ratio in Indonesia and its Relationship with Government Debt and Economic Growth 1990-2022

Thessa Fauziah Pambudi

Tax revenue is the main source of government revenue, which plays an important role in financing national development. The realisation of tax revenue has been increasing during the period from 1990 to 2022. However, the performance of Indonesia's tax revenue as measured by the tax-to-GDP ratio is still relatively low compared to Asia Pacific countries, even its development shows stagnation and decline that it has not been able to meet the international minimum standard of 15 percent. This condition can trigger an increase in public debt and hinder the acceleration of national economic growth. Therefore, this study aims to find out the determinants of tax ratio and its relationship with government debt ratio and economic growth using simultaneous equation model with Two Stage Least Square (2SLS) estimation method. The results show that economic growth, trade openness, and manufacturing share of GDP have positive effect on tax ratio. Tax ratio has a positive effect on economic growth and vice versa economic growth has a positive effect on tax ratio, so there is a two-way positive relationship between the tax ratio and economic growth. Economic growth also has a negative effect on the government debt ratio.

Estimating Conditional Value at Risk in Non-Cyclical Sector Companies Using the Extreme Value Theory Approach

Andi Muhammad Hakam

Conditional Value at Risk (CVaR) is an estimate of the risk of loss that exceeds the Value at Risk (VaR) level. VaR is one of the most commonly used stock risk measurement methods to assess the risk of large investments. Extreme Value Theory (EVT) is a method used to analyze data that contains extreme values. The goal of EVT is to estimate the probability of an extreme event occurring by examining the tails of a distribution based on observed extreme values. There are two general distributions used in EVT, namely Generalized Extreme Value (GEV) and Generalized Pareto Distribution (GPD). This research aims to determine the estimated level of loss that investors may experience when investing in PT Hanjaya Mandala Sampoerna Tbk (HMSP) and PT Japfa Comfeed Indonesia Tbk (JPFA). The L-Moment method is applied to estimate the parameters in this distribution so that an explicit parameter form is obtained. Based on CVaR analysis using the Block Maxima (BM) approach, investors in HMSP and JPFA are estimated to experience losses of 20.0752% and 29.6537% respectively. Using the Peaks Over Threshold (POT) approach, the estimated losses are 0.966% and 1.548% for HMSP and JPFA, respectively. Based on CVaR calculations using both approaches, the POT approach with GPD provides a more accurate and reliable investment risk estimate than the BM approach with GEV distribution.

Use Of Support Vector Machine (SVM), Naïve Bayes And Logistic Regression Methods In User Sentiment Analysis Of Shopee And Lazada Market Place Applications On The Google Play Store Site

Amanatul Khomisah

This research aims to obtain the results of the sentiment classification of shopee and lazada marketplace application users using the support vector machine (SVM), Naïve bayes and Logistic regression methods. Then find out the accuracy results of the support vector machine (SVM), Naïve bayes and Logistic regression methods, and find out the comparison of the best classification results based on the support vector machine (SVM), Naïve bayes and Logistic regression methods in analyzing he sentiment of shopee and lazada marketplace application users. This research uses the support vector machine (SVM), Naïve bayes and Logistic regression methods. The results show that on the Shopee app, SVM,Naïve Bayes, and Logistic regression can classify 489, 492, and 489 reviews with 85,3%, 85,9%, dan 85,3% accuracy, respectively. Meanwhile, on the Lazada app, the SVM, Naïve Bayes, ang Logistic Regression methods can classify 448, 441, 445 reviews with an accuracy 92,6%, 91,1%, 91,9%. The Naïve Bayes method proved to provide the best results for sentiment classification on Shopee with 85,9% accuracy, while the SVM method excelled in sentiment classification on Lazada with 92,6% accuracy

Penerapan Regresi Logistik Biner dan Random Forest Terhadap Pengeluaran Konsumsi Rokok di Kabupaten Gorontalo Tahun 2022

Friansyah Gani

This research discusses classification to determine or predict the class of an object based on available attributes. The purpose of this study is to identify significant factors in cigarette consumption expenditure in Gorontalo Regency using binary logistic regression method as well as to create a classification model using random forest method. The results showed that partially, the education level of the household head, the number of household members and the size of the house played a significant role. However, only the education level of the household head is consistently influential and meets the criteria of goodness of fit in the developed model. On the other hand, in classification analysis, the random forest model showed better performance than binary logistic regression based on the evaluation of classification capabilities such as accuracy, precision, recall, and f1-score. Therefore, random forest was identified as the best classification model in this study.

Optimization of Production Profits with Simplex Method and POM QM Application

Winda Ade Fitriya B

Keripik Keladi Jayapura is one of the UMKM in Jayapura that has been established since 2008. In producing Keripik Keladi Jayapura, business actors have difficulty determining the amount of production of each flavor variant consisting of original flavor (X1), savory flavor (X2), and sweet spicy flavor (X3). In doing production planning, business actors only estimate without being able to know exactly how much of each flavor variant must be produced to be optimal. By setting production priorities so that it can generate maximum profit. The simplex method is used to solve the problem because it has more than two decision variables. In addition to the simplex method, the POM QM application is also used to assist the calculation because POM QM is an alternative software that can help make decisions. Manual calculations using the simplex method produce a profit of 745,000, - if in each production the business actor produces savory flavors (X2) as many as 4 packs, sweet spicy flavors (X3) as many as 21 packs and original flavors (X1) are not produced simultaneously, the calculation results are the same as using the help of the POM QM application.