Vol. 4, No. 1, Januari, 2023 : Estimasi

Livestock in Maluku Province is one of the sectors that is the main priority in the context of increasing people's welfare. The potential for livestock in Maluku Province is increasing every year. However, there needs to be integrated processing and identification of potential commodities in each region. One method that is a reliable statistical method is to use the median clustering method. Median clustering is a method of grouping based on the median value. The median clustering algorithm selects K cluster centers with the aim of minimizing the sum of the measurement distances between each point cluster and the closest cluster center. The data used in this study came from the Maluku Province Central Bureau of Statistics (BPS) in 2022. The results of this research were that there were 3 clusters formed in livestock clusterization in regencies and cities in Maluku Province. Clus ter 1 consists of Southwest Maluku Regency. Cluster 2 consists of the Regencies of Central Maluku, Buru and West Seram. Furthermore, Cluster 3 consists of the Tanimbar Islands, Southeast Maluku, Aru Islands, Eastern Seram, South Buru, Ambon and Tual City.

People's welfare is the goal of the State of Indonesia which is contained in the official state document, namely the opening of the 1945 Constitution paragraph IV, this can be interpreted to enjoy an affluent life, free from poverty and is a human right for every citizen in Indonesia. The grouping process is carried out to see the level of people's welfare for each Regency/City in East Java. In this study, before grouping, the number of clusters was selected using the Elbow method. After that do the grouping with the K-Prototype Algorithm. Furthermore, using Kruskal Wallis and Chi-Square, the test was carried out to determine the variables that influence grouping. The results of the study obtained the 3 best clusters using the Elbow method, grouping with the K-Prototype Algorithm where cluster 1 consisted of 4 Regencies/Cities, Cluster 2 consisted of 18 Regencies/Cities and cluster 3 consisted of 16 Regencies/Cities. Furthermore, the results of Kruskal Wallis and Chi-Square get 4 influential variables in the grouping, the 4 variables are the Number of Poor Population, Expenditures Per Capita, Open Unemployment Rate and Sources of Water for Drinking.

In the regression analysis, the amount of rice production that far exceeds the general production can be categorized as outlier data. The existence of outliers causes the use of the least squares method to estimate parameters to be deemed inappropriate. To deal with outlier data, it is necessary to use methods that are robust or resistant to outlier data. Robust is defined as insensitivity or rigidity to outlier data. The purpose of this study is to obtain a robust regression model using the M estimation, S estimation and MM estimation methods and determine the factors that have a significant effect on rice production in East Nusa Tenggara Province. The model using the S estimation method is the best model, namely y = 3,895.023 + 1.870 X1 - 60.926 X5 and the factors that have a significant effect on rice production are harvested area and air temperature.

Social media is a means to interact with other people through sentences, pictures and videos online. Excessive use of social media has a negative impact on mental health. The grouping process in this study was carried out to see the level of social media use in Bone Bolango Regency. Before grouping, data pre-processing is carried out and the optimal number of clusters is determined using the Silhoutte index. The optimal cluster results obtained are two clusters for all methods. After that, grouping is done using Hierarchical Clustering and Non-Hierarchical Clustering Algorithms. The Hierarchical Clustering algorithm consists of two methods, namely the single linkage method and the complete linkage method. The Non-Hierarchical Clustering Algorithm consists of two methods, namely the K-Means and K-Medoids methods. The next step is to determine the best method using the Davies-Bouldin Index (DBI). The smaller the DBI value, the better the method used. The smallest DBI value is obtained in the complete linkage method. The grouping results for cluster 1 consisted of 70 respondents and cluster 2 consisted of 80 respondents.

The increase of the financial sector, financial information is used in the economy to model and predict the movement of capital market stocks, so investors can easily understand investment risks. Financial sector data is in the form of time series data. Financial data  is found that does not fit the assumption of heteroscedasticity, so a model is needed that can maintain heteroscedasticity. Model Autoregressive Conditional Heteroscedasticity-Generalized Autoregressive Conditional Heteroscedastic is one of the econometric models used to model heteroscedasticity data in time series. The data in this study is BSI's daily closing price data taken from 4 January 2021 to 31 August 2022 with 406 data. Based on the selection of a time series model on Bank Syariah Indonesia (BSI), the best models are ARMA (11.0) and ARCH models (1). So that the ARMA (11.0)-ARCH (1) model can be the best model for modeling and predicting BSI stock return prices.

Detikcom is a very popular news portal today. The news on the portal continues to grow time to time, causing the existing news data to pile up. As a result, this is necessary to utilize this large amount of data. One of the ways that can be used is to extract topics from news text data through topic modeling using the Latent dirichlet allocation (LDA) method. This method is very popular because it can perform analysis on very large documents. This research aims to find certain patterns in a document by generating several different topics so that it does not specifically divide documents into a particular topic. This research has three topics obtained, with a coherence score is 0,7586. The first topic discusses conflicts and crises within a country, the second topic discusses issues related to humanitarian, and the third topic discusses the issues of corruption committed by state officials.

Criminal acts are one indicator of social welfare for a sense of security. The higher the reporting of criminal cases by the public, it indicates that the level of security in the area is getting worse. Crime acts in Tangerang City can be influenced by several factors, namely the poverty factor, the population factor and the population growth rate factor. If the rate of population growth experiences rapid growth, the population will increase and it is undeniable that poverty will increase in the city of Tangerang. This can trigger criminal acts to meet unsatisfied needs. The purpose of this study is to determine the variables that influence criminal acts in Tangerang City and to overcome the variables that occur multicollinearity. It can be concluded that all variables influence crime and the LASSO (Least Absolute Shrinkage And Selection Operator) regression can simplify the model and indirectly overcome the problem of multicollinearity in this study. So that the government can make more efforts to overcome the population and poverty problems that occur and the police to increase security in the City of Tangerang in order to create even better security and minimize crime.

In general, negative binomial regression is used for univariate discrete data that is overdispersive and follows the Poisson distribution. In the real world, a case is often influenced by two discrete variables that are correlated with each other. Therefore, in this paper we will examine the regression that is influenced by two independent variables, has overdispersion properties and follows a bivariate Poisson distribution. This regression is called bivariate negative binomial regression with model parameters estimated using the Maximum Likelihood Estimation (MLE) method and Newton Raphson iterations. The formation of this model is based on the Famoye method, while in general it uses the Cheon method. Furthermore, the results of this study were applied to data on the number of maternal and infant deaths in South Sulawesi Province in 2020. The results obtained were the number of puskesmas that had a significant effect on the number of maternal deaths and the proportion of handling obstetric complications, the proportion of pregnant women implementing the K4 program, the proportion of deliveries in facilities health services, the proportion of postpartum mothers implementing the KF2 program and the number of puskesmas have a significant effect on the number of infant deaths.

Covid-19 in Indonesia began to be recorded on March 2, 2020 with the number of positive patient cases as many as 2 people with the passage of time Covid-19 cases in Indonesia are always increasing. To see the development of Covid-19 cases in the future period, the opportunity for the number of Covid-19 cases can be used using the Markov chain. The Markov chain method is carried out using a transition probability matrix which is seen from the number of additions to positive Covid-19 patients in a steady state or a situation for a long period of time. Based on the results of the range of additions to the number of positive cases of Covid-19, 6 states were used. Furthermore, the calculation of the Markov Chain in the stationary state of Covid-19 cases in Indonesia after 328 days or 11 months obtained the probability of each state, namely state 1 of 0.0005, state 2 of 0.0069, state 3 of 0.1707, state 4 of 0.1462, state 5 of 0.1884 , and state 6 is 0.4873. Prediction of the addition of positive Covid-19 patients obtained results as many as 2058 patients in state 5 for July 1, 2022 with actual data as many as 2049 patients.

In the Mixed Geographically Weighted Regression (MGWR) model, some variables are local and some are global. In MGWR modeling, it is often found that the data have multicollinearity. To overcome this problem, MGWR models with ridge regression are used. The MGWR model can be applied to poverty cases because it can experience spatial heterogeneity due to differences in geographical, cultural, and economic policies that vary in each region. In this study, the estimation of MGWR model parameters with ridge regression is then applied to data on the poor population of South Sulawesi in 2016. Data on the poor population of South Sulawesi experience multicollinearity, so it is solved using the MGWR model with ridge regression. Variables that have a significant effect globally are x3  and x6. while the variables that have a significant local effect are x2, x4, x5, x7, x8, x9 and x10. The AIC value of the MGWR model with ridge regression of 63.64473 is smaller than the MGWR model, meaning that the addition of ridge regression to the MGWR model makes the model better at overcoming multicollinearity problems.

Overdispersion is a state with a variance value greater than the mean value so the Poisson Inverse Gaussian regression model is used. Meanwhile, to model two correlated response variables, the Bivariate Poisson Inverse Gaussian (BPIG) regression model was used. The BPIG model is a mixed- distributed model between the Poisson Bivariate and Gaussian Inverse distributions. The parameters of the BPIG regression model are estimated using Maximum Likelihood Estimation (MLE) with the Fisher Scoring algorithm. This study was applied to data on the number of maternal and neonatal deaths in South Sulawesi in 2019. The results obtained are predictor variables that affect the number of maternal and neonatal deaths in South Sulawesi in 2019, namely K4 services for pregnant women , active birth control participants , handling obstetric complications , handling neonatal complications  and the number of health centers .

One of the assumptions that must be met in the multiple linear regression model is that there is no multicollinearity problem among the independent variables. However, if there is a multicollinearity problem, then parameter estimation can be done using the linearized ridge regression (LRR) method. The LRR method has the advantage of choosing an optimal constant that is easy to determine and also has a minimum PRESS value. In this study, the infant mortality rate in South Sulawesi Province will be modeled using the LRR method based on the variables of the amount of vitamin A given, the number of health services, the number of babies born with low weight, the number of mothers who give birth assisted by medical personnel, and the number of babies who are breastfed. exclusive. One measure to see the goodness of the regression model is the Prediction Error Sum of Squares (PRESS). Based on the t-test at a significance level of 5%, the total coverage of vitamin A administration and the number of babies born with low weight gave a significant effect on infant mortality with a PRESS value of 0.6846.