EKSPONENSIAL

Mengatasi Multikoliniearitas Dalam Regresi Linier Berganda Menggunakan Principal Component Analysis

Niken Harel Chairunnisa — 2025-04-17

Multiple linear regression analysis has assumptions that must be met, one of which is multicollinearity. Multicollinearity occurs when the independent variables correlate with each other, resulting in the regression coefficient produced by multiple linear regression analysis being very weak or unable to provide analysis results that represent the nature or influence of the independent variable concerned. The detection of multicollinearity can be known through the VIF value. In this study, human development index data on Kalimantan Island in 2019 detected multicollinearity because some independent variables have a VIF value of more than 10 so that the method used to overcome multicollinearity in this study is Principal Component Analysis (PCA). Based on the results of research using the Principal Component Regression method, There are five independent variables that influence the IPM that is Percentage of Poor Population, Number of Health Workers, Number of Workforce, Number of High Schools, and Number of High School Teachers.

Perbandingan Regresi Robust dengan M, S, dan MM-Estimator untuk Menganalisis Faktor-Faktor yang Memengaruhi Indeks Pemberdayaan Gender di Nusa Tenggara Barat Tahun 2023

Syfriza Davies Raihannabil — 2025-04-17

The government has targeted gender issues in the fifth sustainable development goal, one of which is to achieve gender empowerment. The indicator used to measure gender empowerment in Indonesia is the Gender Empowerment Measure (GEM). NTB has been the province with the lowest GEM in Indonesia for five consecutive years, from 2019 to 2023. In addition, in 2023, NTB experienced a decrease in GEM of 0.19 points from 2022. This research aims to analyze factors that influence GEM in NTB in 2023. However, outliers are often found in the data which makes estimates using OLS biased. Therefore, this research uses a robust regression analysis method to overcome outliers in the data by comparing parameter estimates between M, S, and MM-estimator. The analysis results show that the best estimation method is the S-estimator because it produces the highest and the lowest residual standard error (RSE) between the M and MM-estimator. All predictor variables have a positive and significant effect on GEM, namely women's involvement in parliament , women as professionals , and women's income contribution . The S-estimator produces a of 0,999, which means that all predictor variables used can explain a proportion of GEM diversity of 99,9%, while the remainder can be explained by other variables that are not included in the model.

Analisis Klaster Menggunakan Metode Average Linkage dengan Validasi Multiscale Bootstrap (Studi Kasus: Indikator Pendidikan di Indonesia Tahun 2021)

Fikri Rasidia — 2025-04-17

The average linkage method is one of the hierarchical cluster analyses, where the clustering process starts by finding two objects that have the closest distance to the average rule of the two groups. The multiscale bootstrap method is a method used to see the validity of the cluster analysis results. This study aims to determine the clusters formed using the average linkage method, as well as to determine the validity of the clusters formed based on education indicators in each province in Indonesia. The result of the study is one cluster with AU (Approximately Unbiased) ≥ 0.95 so that the cluster is considered to be able to represent the actual population.

Analisis Potensi Pencemaran Air Sungai Di Lingkungan Hutan Tropis Lembap Kalimantan Timur Menggunakan Model Regresi Weibull

Yola Desty — 2025-04-17

Weibull Regression Model (WR) is the Weibull distribution in which scale parameter is stated in the regression parameter. WR model derived from the interrelated functions of Weibull Distribution, consisting of Weibull survival regression model, Weibull cumulative distribution regression model, Weibull hazard regression model, and Weibull mean regression. The purpose of this study was to obtain the pollution potential information of river water in east Kalimantan and to obtain the factors that influence it through RW modeling on dissolved oxygen (DO) data in 2022. Research data is secondary data provided by life inveronment of East Kalimantan Province. The parameter estimation method is maximum likelihood estimation (MLE). The study concluded the pollution potential information of river water in east Kalimantan Timur based on modeling RW DO data consists of the chance the unpolluted river water is 0.6868, chance of polluted river water is 0.3132, the water pollution rate is 0.4349 locations/ ppm, and average river water DO is 5.6003 ppm. Factors that influence the pollution potential of river water is nitrite concentration, water temperature, and degree of water color.

Penerapan Spatial Error Model (SEM) Dalam Menganalisis Faktor-Faktor Yang Mempengaruhi Stunting Balita Di Indonesia

Zakiyah Mar'ah — 2025-04-17

Stunting, a major public health concern hindering child development, remains prevalent in Indonesia. This study employs a spatial approach to analyze the prevalence and spatial patterns of stunting across 34 provinces in Indonesia in 2022. We utilize Exploratory Spatial Data Analysis (ESDA) with Moran's I to assess spatial autocorrelation and identify potential model types (e.g., Spatial Autoregressive Model (SAR), Spatial Error Model (SEM), General Spatial Model (GSM). Following this, Local Indicators of Spatial Association (LISA) can be employed to pinpoint specific spatial clusters of high or low stunting prevalence. The analysis confirms spatial autocorrelation, and subsequent modeling using a suite of spatial regression techniques (including SAR, SEM, and SARMA/GSM) reveals the SEM as the most suitable model for this study with the weighting of the queen matrix contiguity. The SEM analysis identifies two key factors influencing stunting rates: the percentage of the poor population and the percentage of infants under 6 months receiving exclusive breastfeeding. This study highlights the importance of a spatially informed approach for developing effective national and regional stunting prevention programs. By targeting interventions in provinces with high stunting clusters and addressing underlying factors like poverty and breastfeeding practices, policymakers can create more equitable resource allocation strategies to combat stunting and improve child health outcomes nationwide.

Model Regresi Spasial pada Proporsi Tenaga Kerja Perempuan di Provinsi Sulawesi Selatan

Zakiyah Mar'ah — 2025-04-17

Female labor force participation in South Sulawesi, Indonesia, is an urgent issue in the context of economic development and gender equality. For this issue, spatial regression is performed to build the relationship between variables that influence female labor force participation in the region. This study performed the Spatial Autoregressive (SAR) model, which is a regression model where the response variable has spatial correlation. The value of Moran's I for the proportion of female labor force in South Sulawesi is 0.05125, meaning there is a positive spatial autocorrelation. The results obtained showed that the expected length of schooling and adjusted per capita expenditure have a positive effect and the average length of schooling has a negative effect on the proportion of female labor force in South Sulawesi.

Klasifikasi Naïve Bayes Pada Data Status Kesejahteraan Rumah Tangga Penerima Manfaat di Kecamatan Samarinda Ilir Tahun 2023

Indah Cahyani Lupinda — 2025-06-03

Data mining is the process of extracting useful information and patterns from very large amounts of data. Based on the task or work performed, data mining is divided into cluster analysis, association analysis, anomaly detection, and predictive modeling. Predictive modeling consists of two types, namely regression and classification. Classification is a method for determining the membership of an object in a class based on available data. There are several methods for classification, one of which is naïve Bayes with the advantages of being easy to build and having good performance. This research aims to determine the results of the accuracy of the naïve Bayes classification on data on the welfare status of beneficiary households in Samarinda Ilir District in 2023. Based on the research results, it can be seen that the accuracy level of the naïve Bayes classification on this data is 0.8316 or 83.16%. The results of accuracy measurements show that the naïve Bayes classification of this data has a fairly high level of accuracy.