UNP Journal of Statistics and Data Science

Modeling of Employment Participation Rate Against the Percentage of Poor Population in East Java in 2023 Using B-Spline Method

2024-08-12T06:25:22+00:00

Poverty is a common issue in Indonesia. Data on the Percentage of Poor Population against the Labor Force Participation Rate (LFPR) per district/city, consisting of 38 districts/cities in East Java Province in 2023, indicates that the highest percentage of poverty in East Java Province in 2023 was 21,760. Employment is considered the most effective solution to alleviate poverty. The data in this study shows a distribution pattern that does not form a specific pattern, making it difficult to analyze using parametric methods. Therefore, the appropriate approach is Nonparametric Regression. In this study, the nonparametric regression used is the B-Spline regression model. The suitability of the model is based on the Mean Squared Error (MSE) value of the model. The analysis results indicate that the B-Spline regression model achieves an MSE value of 20.11447. The optimal MSE value is obtained from B-Spline estimation with order 2. This suggests that the B-Spline method provides a good explanation in addressing the issue

Estimation of Poverty in North Sumatera in 2022 using Truncated and Penalized Spline Regression

2024-08-17T03:05:25+00:00

The Sustainable Development Goals' main goal is to reduce poverty (SDGs). Low human capital is the cause of poverty. The Human Development Index is one indicator that can be used to assess human capital (HDI). Despite having the largest population on the island of Sumatra, North Sumatra continues to have the fifth highest poverty rate. Because the pattern of the relationship between poverty and HDI based on previous research is still unclear because the results are inconsistent, nonparametric regression modeling was used in this study because it is flexible in following the pattern of data relationships and can avoid model prespecific errors. This study aims to compare the Spline Truncated and Penalized Spline regression methods. The results of the comparison between the Truncated Spline regression model and the P-Spline regression model by looking at the smallest MSE value showed that a better estimator for modeling the Human Development Index in North Sumatera in 2022 is non-parametric regression using the truncated spline estimaor. where the best truncated spline modeling is at order 2 with one knot point located at X = 66.93 with a GCV value of 6.0543.

Optimization of Sentiment Analysis for MBKM Program using Naïve Bayes with Particle Swarm Optimization

2024-08-12T07:13:43+00:00

In early 2020, Kemendikbudristek launched the MBKM program with the aim of improving the quality of higher education through a student-focused learning approach. The launch of this program triggered various reactions on social media, especially on Twitter, both positive and negative. This study aims to analyze the sentiment of Twitter users towards the MBKM program using the Naive Bayes algorithm optimized with Particle Swarm Optimization (PSO). The data used are Indonesian tweets containing the keywords "MBKM" and "Merdeka Campus" from the period July to December 2022. The research stages include data collection through crawling, manual labeling of data into positive and negative sentiments, data preprocessing, application of the Naive Bayes algorithm, and feature selection with PSO. The results showed that the group of tweets categorized based on positive and negative sentiments towards the implementation of the MBKM program in Indonesia in 2022, showed that the NB-PSO experiment achieved an accuracy of 90.87%, an increase of 7.12% compared to the Naive Bayes algorithm alone. Thus, the use of Particle Swarm Optimization algorithm in Naive Bayes classification algorithm is proven to improve classification performance, especially in the case of sentiment analysis.

Keywords: Sentiment Analysis, Merdeka Belajar Kampus Merdeka, Twitter, Naive Bayes, Particle Swarm Optimization.

Application of Choice-Based Conjoint Analysis Method in Students Job Preference at Statistic Department of Universitas Negeri Padang

2024-11-15T09:16:03+00:00

In the realm of psychology studies, it is widely assumed that the age range between 18 and 25 represents a critical period during which individuals preferences begin to take shape. This developmental phase encloses college students who despite their academic pursuits, remain relatively unfamiliar with the dynamic job market, particularly in the context of rapid technological advancements. Statistics as a discipline with broad applicability across both social and scientific domains, offers student of statistics significant career prospects. This research would likely estimate the job preferences of statistics students using one of the most common use methods called choice-based conjoint (CBC) analysis. The analysis reveals that work hours were the most substantial influence on statistics students’ job preferences, with a percentage of 40.29%. In addition, other factors that influence the preferences of statistics students are such as first salary (36.87%), correlation with the field of statistics (12.04%), work environment (7.18%), and type of workplace (3.62%).

Application of Rating-Based Conjoint Analysis Method in Students E-Wallet Preference at Statistic Department of Universitas Negeri Padang

2024-11-15T09:16:23+00:00

The rapid development of technology in the era of globalization has influenced the evolution of society's life in terms of economy, social, culture, and education, with the aim of facilitating daily activities, one of which is the ease of transactions using e-wallets. An e-wallet is a payment tool that uses a server-based system. Many factors influence a person's decision to use an e-wallet as a payment method, one of which is the level of security. To identify the factors that affect someone's use of e-wallets, one method is Rating-Based Conjoint Analysis (RBC). Therefore, this study aims to determine what influences a person to use an e-wallet, with the subjects being active students of the Statistics Department at Padang State University. The results of this RBC study indicate that the most influential factor on the e-wallet preferences of statistics students is security level, with a value of 37.70%, followed by transaction speed 23.17%, transfer fees at at 23.07%, features provided at 11.78%, and the least influential factor being promotions at 4.28%.

PT.Telkom (Tbk) Stock Price Forecasting Using Long Short Term Memory (LSTM)

2024-08-12T07:24:38+00:00

The movement of the share price of PT Telkom (Tbk) fluctuates so it is necessary to do a forecasting analysis. Forecasting the share price of PT Telkom (Tbk) can be done using the Long Short Term Memory (LSTM) method. LSTM is a development of the Recurrent Neural Network (RNN) method. In this study using PT.Telkom (Tbk) stock price data for 2018-2023 and PT.Telkom (Tbk) stock price data after Covid-19 (20121-2023). The purpose of this research is to determine the movement of PT.Telkom (Tbk) stock prices in 2024, to find out the difference in forecasting using PT.Telkom (Tbk) 2018-2023 stock price data with PT.Telkom (Tbk) stock price data after covid-19 2021-2023, and to determine the level of accuracy of forecasting PT.Telkom (Tbk) stock prices using the LSTM method. The results showed that both data have a small MAPE value. to forecast the share price of PT.Telkom for 1 year, PT.Telkom (Tbk) share price data for 2018-2023 is used which has more data to analyze long-term forecasting. From the analysis results obtained MAPE of 1.016% with the optimal parameter combination of neuron 4, batch size 64, and epoch 80. The results of forecasting the share price of PT.telkom (Tbk) in 2024 experienced very rapid fluctuations with an average share price of PT.Telkom (Tbk) in 2024 Rp 4,668 / sheet.

Implementation of CART Method with SMOTE for Household Poverty Classification in Mentawai Islands 2023

2024-11-09T10:23:30+00:00

Poverty is a condition in which individuals or groups are unable to fulfill their basic needs due to economic pressure or limited resources. The Classification and Regression Trees (CART) method is a classification technique in the form of a classification tree, which describes the relationship between independent and dependent variables. Data imbalance can lead to low sensitivity values and area under curve (AUC) values. One method that can overcome unbalanced data is to perform Synthetic Minority Oversampling Technique (SMOTE). SMOTE is a technique with the addition of artificial data in the minority class at a stage before analyzing the data. The purpose of this research is to compare the model without and with SMOTE in CART method. The use of SMOTE is applied to balance the amount of data on each poor household. The accuracy value of the method without SMOTE is 89% while with the SMOTE method is 79%. However, the sensitivity value has increased by 80%. Meanwhile, the AUC value in the CART method with SMOTE increased by 31%. So in this study it can be concluded that CART classification analysis with SMOTE is able to provide better performance compared to CART classification analysis without SMOTE.

Factors Influencing the Number of Families at Stunting Risk in Merangin Regency Using Mixed Geographically Weighted Regression

2024-08-13T01:34:15+00:00

The number of families at risk of stunting is among the significant concerns that have been a negative impact on developing superior human resources in Merangin Regency. The number of families at risk of stunting is sought to be solved by identifying the contributing components. MGWR is among the methods that may be employed to obtain a specific model that affects each obesrvasion location locally and a comprehensive model that is global. Multiple linear regression and GWR are used to create models MGWR used when data has the influence of spatial heterogeneity. This project aims to develop an MGWR model which will be used to calculate the amount families at risk of stunting in each sub-district in Merangin Regency who are at risk of stunting in 2022. A fixed gaussian kernel weighting matrix is used in MGWR modeling. At the very least CV of 0.6152241, A fixed gaussian kernel is utilized as the weighting function. The results indicate that the model obtained has an accuracy rate of 99.18%, which means that the predictor variables can explain the model by that percentage. Families with insufficient access to drinking water is one factor that significantly affects how many families are at risk of stunting, families with inadequate sanitation, maternal age less than 20 years and families with babies under five years old.

Early Marriage Factors Indonesian Using Spatial Regression Analysis

2024-08-13T01:35:07+00:00

Marriage is a sacred union recognized socially and religiously to form a family, as regulated by Law No. 16 of 2019. The percentage of early marriages in Indonesia continues to rise, reaching 21.5% in 2022, placing Indonesia 8th in the world according to UNICEF 2023 data. The increase in early marriages has significant impacts on maternal and child health and often leads to high divorce rates, with 516,334 cases in 2022. The aim of this research is to provide information and knowledge for students about early marriage and spatial regression. The main factors influencing early marriages are low education levels, economic difficulties, and environmental factors. Research shows that early marriages are highest in Kalimantan and Sulawesi, with spatial effects influencing the percentage of early marriages between regions.Spatial regression analysis, such as the Spatial Autoregressive (SAR) model, is used to examine the interactions between regions affecting early marriage. Spatial autocorrelation tests and spatial dependency effects show a spatial dependency effect, making the SAR model with queen contiguity weights the most suitable. The resulting model is considered quite good considering the R-squared value of 40.97%. The best-formed model shows that the Open Unemployment Rate (TPT) of youth is a significant variable that greatly impacts the percentage of early marriages. Therefore, the central and provincial governments are expected to pay more attention to the open youth unemployment factor to control and reduce the rate of early marriages in Indonesia.

Classification of Poor Households in Padang City Using the Naïve Bayes Algorithm with Synthetic Minority Oversampling Technique

2024-08-13T01:36:05+00:00

Poverty is a condition where a person is unable to meet minimum basic needs or a condition caused by the influence of development policies that have not been able to reach all levels of society. In Indonesia, the government has designed various programs to overcome poverty, but these programs are often not on target. One method to improve the effectiveness of the program is through proper classification of poor and non-poor households. This study uses the Naïve Bayes classification method which is popular in data mining to predict data categories based on the probability distribution of its features. However, challenges arise when the data is unbalanced between different classes. To overcome this, the Synthetic Minority Oversampling Technique (SMOTE) method is used to balance the data. Based on the analysis that has been carried out To determine the performance of Naïve Bayes using SMOTE and without SMOTE in classifying poor households in Padang City in 2023, classification using the Naïve Bayes method without SMOTE produced an accuracy value of 98%, precision of 0%, and recall of 0%. Meanwhile, the classification using the Naïve Bayes method with SMOTE produces an accuracy value of 90%, precision of 87%, and recall of 92% and the results of the criteria for poor households in Padang City in 2023 using Naïve Bayes can be seen from the results that the probability of poor households is much greater than that of non-poor households, therefore the data is classified as group of households that are classified as poor.

Library Book Lending Recommendation Using Association Rules with Frequent Pattern Growth (FP-Growth) Algorithm

2024-11-04T02:57:57+00:00

College libraries are libraries managed by higher education institutions such as university libraries. The library functions as an information center management forum for students which includes learning resource functions, access functions, librarian functions, ethical functions, and evaluation functions. Students prefer to read through e-books rather than reading books or library collections. Limited knowledge of literature is the cause of students choosing to look for books on search engines rather than in the library. Managed book loan circulation history data will be able to improve library services that can assist in finding library collections. Book recommendation services using association rules, can find patterns of borrowing behavior of book titles that have the highest association as the most recommended titles to be borrowed together. The FP-Growth or Frequent Pattern Growth is an algorithm of associations rule that is able to generate association rules as personalized book borrowing recommendations. The results of book recommendations found as many as 50 rules that meet the chi-square assumption test where the recommendation items are independent. The results of 50 rules for book title choices that can be used by students as suggestions for determining books that have a relationship to be borrowed together to enrich references. For students who wish to borrow the books 'Professional Teacher: Mastering Teaching Methods and Skills' is recommended to also borrow the book 'Participatory Learning Methods and Techniques'. With the book recommendation service, the library provides advice to students in choosing related book titles to borrow at the library.

Prediction of World Gold Price Using k-Nearest Neighbor Method

2024-11-06T03:34:00+00:00

This research aims to predict world gold prices using the k-nearest neighbor (KNN) method with secondary data from the London Bullion Market Association (LBMA) in the form of monthly time series data from January 2019 to December 2023. In the analysis process, the data is divided into two parts: 80% for training data (January 2019 - December 2022) and 20% for testing data (January - December 2023). The analysis results show that the Mean Absolute Percentage Error (MAPE) value of the KNN method is 4.5%, which indicates a very good level of accuracy. With a MAPE below 10%, the KNN model is proven to be able to accurately predict world gold prices. Gold price predictions for the period January to December 2024 show a consistent upward trend, which is influenced by factors such as global economic fluctuations, increased gold demand, and geopolitical uncertainty. These results show that the KNN model is reliable as a tool for forecasting future world gold prices.

Comparison of Quadratic Discrimination Analysis with Robust Quadratic Discrimination

2024-11-06T03:37:46+00:00

This study compared the performance of quadratic discrimination analysis and robust quadratic discrimination analysis using the Iris dataset from Kaggle. The robust quadratic discriminant analysis, designed to handle outliers and non-normal distributions, shows better performance with an Apparent Error Rate (APER) of 2.5%. In contrast, the quadratic discriminant analysis, used for data with multivariate normal distribution and different variance-covariance matrices among groups, yields an APER of 3.03%. These results indicate that robust quadratic discriminant analysis is more accurate in classification on this dataset compared to quadratic discriminant analysis.

Keywords: Apparent Error Rate, Quadratic Discrimination Analysis, Robust Quadratic Discrimination Analysis

Regularized Ordinal Regression with LASSO: Identifying Factors in Students' Public Speaking Anxiety at Universitas Negeri Padang

2024-11-06T04:14:23+00:00

Public speaking anxiety is a common issue faced by students, particularly in academic settings. It may arise from a range of factors, including humiliation, physical appearance, preparation, audience interest, personality traits, rigid rules, unfamiliar role, negative result, and mistakes. This research seeks to determine the factors influencing different levels of public speaking anxiety among students at Universitas Negeri Padang through the application of ordinal regression with LASSO regularization. This method allows for automatic selection of significant variables and addressesmulticollinearity issues. The results indicate that eight factors influence low public speaking anxiety levels, while only six factors impact high public speaking anxiety levels. The ordinal regression model with LASSO penalty demonstrates good performance in classifying public speaking anxiety levels, achieving an accuracy of 71.33%. This study is expected to help students and educators better understand and manage public speaking anxiety, thereby enhancing public spekaing competence among students

Comparison of Naïve Bayes and K-Nearest Neighbors Methods in Classifying Human Development Index by Districts/City Indonesia in 2022

2024-11-15T09:21:06+00:00

The Human Development Index (HDI) is an indicator used to measure the success of efforts to improve the quality of human life in a particular region. Indonesia's HDI has increased every year, but the HDI in several districts/cities in Indonesia remains in the low category. The low HDI in these districts/cities is due to unequal development between regions in Indonesia. This disparity in development is influenced by HDI indicators as well as other factors. To address this issue, a decision system is needed to determine HDI categories using the Naive Bayes and KNN methods. Naive Bayes is applied with the assumption of Gaussian distribution, while KNN is implemented with the optimization of the nearest K value. Model performance evaluation is conducted to determine the best accuracy of the two methods using a confusion matrix. The analysis results show that the Naïve Bayes model outperforms the KNN algorithm in classifying the Human Development Index (HDI) by district/city in Indonesia for the year 2022, with Naïve Bayes achieving an accuracy of 93%. Therefore, the Naïve Bayes algorithm show good performance in terms of accuracy.

Sentiment Analysis of The Constitutional Court Decision Regarding Changes to The Age Limit for Presidentian and Vice Presidential Candidates Using Support Vector Machine

2024-11-06T03:54:32+00:00

The Constitutional Court (MK) as a judicial institution granted a judicial review on October 16, 2023 related to the Election Law Article 169 (q) Law No.7 of 2017 number 90/PUU-XXI/2023. The Constitutional Court approved the material test, leading to changes in the age limit for presidential and vice presidential candidates. This change caused controversy because it was considered to benefit one of the candidate pairs. This research aims to see the trend of public opinion towards policy changes by the government. This research uses the Support Vector Machine (SVM) method which divides the data into two classification classes. The application of linear, Radial Bias Function (RBF), and polynomial kernels resulted in the highest accuracy of 84%. The calculation of accuracy, precision, and recall is 84%, 22%, and 90%, respectively. Based on the resulting wordcloud, Positive words indicate backing for presidential and vice presidential candidates. Meanwhile, negative sentiments express disapproval of the Constitutional Court's decision concerning the changes to the age limit requirements for presidential and vice presidential candidates.

How MUI Fatwa Changes Indonesia Mindset towards Pro-Israel Boycott Products using the Naïve Bayes Classification Method

2024-11-27T13:48:19+00:00

Boycotting pro-Israel products has become a popular topic on social media, both in Indonesia and globally. This research aims to analyze the sentiments of Indonesian using the Naive Bayes classification method regarding the boycott before and after the issuance of MUI Fatwa No.83/2023. Through sentiment and word cloud analysis of 3327 tweets, it was found that discussions remained consistent and were not influenced by MUI Fatwa. The sentiment of the majority of Indonesian regarding the boycott of pro-Israel products is positive, with full support for this action. MUI Fatwa has had an impact on the sentiment of Indonesian, as can be seen from the increase in positive sentiment after the fatwa was released. Word cloud analysis shows that both before and after November 8, 2023, the top one word that appears in the word distribution is exactly the same, namely 'boycott'. This similarity shows that the discussion topics that developed on the Twitter platform remained consistent, both before and after the release of MUI Fatwa Indonesian netizens have uniformly discussed boycotting products that support Israel as a form of rejection of the genocide carried out by that country in Gaza, Palestine.

Mapping Indonesian Provinces Based on Leading Plantation Commodities with Export Potential Using Multidimensional Scaling Analysis

2024-11-27T13:49:04+00:00

Indonesia, as an agrarian country, benefits significantly from its plantation subsector, which contributes substantially to the national economy. However, the processing of plantation products in Indonesia remains largely limited to raw or semi-finished goods, resulting in low added value and restricted income for both farmers and the nation. This study aims to map Indonesia's provinces based on the production of key plantation commodities with high export potential, utilizing the Multidimensional Scaling (MDS) analysis method. The research focuses on commodities such as pepper, palm oil, coconut, rubber, coffee, cocoa, clove, and tea. It seeks to group 34 Indonesian provinces based on similarities in plantation production, providing valuable insights for policymakers to enhance production and increase export value. The analysis calculates inter-provincial similarities to determine distances between objects and evaluates the accuracy of the MDS mapping using STRESS and R² values. The findings indicate that 12 provinces share similarities in cocoa production, while 7 provinces are closely aligned in the production of pepper, rubber, and coffee. Furthermore, 5 provinces exhibit similarities in palm oil production, and 9 provinces demonstrate commonalities in the production of coconut, clove, and tea. The analysis achieved a STRESS value of 0.024 (2.4%) and an R² value of 0.9994, indicating that the MDS mapping is highly reliable. However, the results do not fully align with field data, suggesting the need for orthogonal transformation through Principal Component Analysis (PCA) to improve accuracy.