FORUM STATISTIKA DAN KOMPUTASI
Not a member yet
119 research outputs found
Sort by
PENGKLASIFIKASIAN KOLEKTIBILITAS NASABAH KREDIT MENGGUNAKAN METODE SIMPLE NAÏVE BAYESIAN CLASSIFIER
No full textDalam banyak kesempatan, penyusunan model skoring untuk memprediksi klasifikasi calon nasabah dilakukan menggunakan model regresi logistik dan beberapa model lain. Proses pengklasifikasian dapat juga dilakukan dengan menerapkan simple naive Bayesian classifier. Meskipun menggunakan asumsi yang secara umum dilanggar oleh data dan proses komputasi yang jauh lebih sederhana, teknik ini mampu menghasilkan akurasi dugaan yang tidak mengecewakan. Paper ini memberikan ilustrasi penggunaan simple naive bayesian classifier pada kasus prediksi klasifikasi status kolektibilitas calon nasabah dan membandingkannya dengan model regresi logistik dan generalized additive model. Kata kunci: simple naive Bayesian classifie
PENDEKATAN KEKAR UNTUK MODEL BERSAMA (JOINT MODEL) ATAS DASAR SEBARAN t (A Robust Approach for Joint Model Based on t Distribution)
Full text linkExisting methods for joint modeling are usually based on normality assumption of random effects and intra subject errors. We propose a joint model based on t distribution of the intra subject errors to improve robustness of the estimation. Our model consists of two submodels: a mixed linear mixed effects model for the longitudinal data, and a generalized linear model for continuous/binary primary response. The proposed method is evaluated by means of simulation studies as well as application to HIV data. Keywords: joint modeling, longitudinal data, robust, t distributio
THE LOG LINEAR MODELS FOR TWO DIMENSIONAL CONTINGENCY TABLES UNDER THE MULTINOMIAL SAMPLING DESIGN
Full text linkNegative binomial regression model is used to overcome the overdispersion in Poisson regression model. This model can be used to model the relationship of the infant mortality and the factors incidence. Geographical conditions, socio cultural and economic differ one of location another location causes the factors that influence infant mortality is different locally. Geographically Weighted Negative Binomial Regression (GWNBR) is one of methods for modeling that count data have spatial heterogeneity and overdispersion. The basic idea of this model considers of geography or location as the weight in parameter estimation. The parameter estimator is obtained from Iteratively Newton Raphson method. This research will determine the factors that influence infant mortality. GWNBR model with a weighting adaptive bi-square kernel function classifies regency/city in East Java into 16 groups based on the factors that significantly influence the number of infant mortality. This model is better used to analyze the number of infant mortality in East Java in 2008 due to a smallest deviance value.Keywords : Negative binomial regression, geographically weighted negative binomial regression, adaptive bi-square, overdispersio
PENGGUNAAN JARINGAN SYARAF TIRUAN UNTUK PENDUGAAN MODEL LINEAR TERAMPAT DENGAN KOEFISIEN KERAGAMAN KONSTAN
Full text linkSecara umum model linear terampat (GLIM) dapat dipetakan secara ekivalen pada Jaringan Syaraf Tiruan (JST) dengan satu lapisan atau disebut juga dengan perceptron. Fungsi aktifasi pada JST sama dengan invers dari fungsi hubung. Pada GLIM dengan komponen acak mempunyai sebaran gamma ekivalen dengan JST tanpa lapisan tersembunyi dengan fungsi galat adalah gamma dan fungsi tujuan adalah fungsi kemungkinan maksimum atau devians. Sedangkan fungsi aktifasi untuk model gamma adalah identitas, resiprokal, atau eksponensial. Makalah ini mengkaji pendugaan model pada data yang mempunyai sebaran gamma dengan metode JST dan seberapa besar perbedaan hasil pendugaannya dibandingkan dengan GLIM. Hasil kajian menunjukkan bahwa JST menghasilkan nilai dugaan yang sama dengan GLIM. Kata kunci : jaringan syaraf tiruan, koefisien keragaman konstan, model gamm
PERBAIKAN METODE KRIGING BIASA (ORDINARY KRIGING) MELALUI PEMECAHAN MATRIKS C MENJADI BEBERAPA ANAK MATRIKS NON OVERLAP UNTUK MEWAKILI DRIFT PADA PEUBAH SPASIAL
Full text linkPersoalan dalam pendugaan spasial dengan menggunakan konsep drift sering kali menemui kendala bila kondisi permukaan yang diduga bersifat anisotropik. Pada kondisi anisotropik kuranglah tepat apabila hanya menggunakan satu model korelogram (variogram). Dalam tulisan ini dicoba area yang diteliti dibagi menjadi beberapa partisi (4 partisi) sehingga disusun empat model variogram untuk keseluruhan area. Dari empat partisi tersebut dicari nilai-nilai total pembobot yang layak agar fungsi penduga menjadi tak bias. Selanjutnya dilakukan perbandingan pendugaan nilai pada titik-titik yang tidak dilakukan pengukuran antara tanpa partisi dengan partisi. Hasil pendugaan menunjukkan bahwa nilai dugaan sama dengan nilai sebenarnya baik yang tak dipartisi maupun yang dipartisi. Akan tetapi, nilai pendugaan yang dihasilkan dari area yang dipartisi lebih baik dibandingkan tanpa partisi
A SUMMARIZATION OF BAYESIAN ASPECTS OF SMALL AREA ESTIMATION
No full textBayesian approach in Small Area Estimation (SAE), namely Empirical Bayes (EB) and Hierarchical Bayes (HB), will be described. EB approach in SAE has been given considerable attention by some practitioners because it does not require specification of the prior distribution for model parameters and it is usually obtained in a closed form. However, an accurate measure of uncertainty of EB estimator can not be obtained unlike HB estimator in which its uncertainty can be measured exactly. On the other hand, HB approach requires subjective prior distribution for model parameters and it is usually not easy to be specified in practice especially for public policy. Keywords : small area estimation, empirical bayes, hierarchical baye
PENGENALAN ALGORITMA GENETIK UNTUK PEMILIHAN PEUBAH PENJELAS DALAM MODEL REGRESI MENGGUNAKAN SAS/IML
Full text linkGenetic algorithm has been a popular alternative in the various fields of optimization problem. This paper describes some basic ideas of this algorithm and its application for selecting significant variables in the regression analysis. Simple SAS/IML commands are presented in order to emphasize how the algorithm works. It is also available to do some modification in some parts of those commands
PEMILIHAN MODEL REGRESI LINIER MULTILEVEL TERBAIK
Full text linkLinear regression models is used to describe relationship between dependent variable and independent variables. In a survey research, data was used often have hierarchical structure or nested structure. In this research, independent variables can be defined at any level of the hierarchy but dependent variable can only be defined at the lowest level of the hierarchy. Multilevel regression models is one of the methods can be used to analyze this data. Some authors purpose many models can be used to analyze data with hierarchical structures. Deviance as -2 log likelihood was defined as the measure goodness of fit. The difference of the deviance for two nested models was a method for comparing that two models
MODEL REGRESI BINOMIAL NEGATIF TERBOBOTI GEOGRAFIS UNTUK DATA KEMATIAN BAYI (Studi Kasus 38 Kabupaten/Kota di Jawa Timur) (Geographically Weighted Negative Binomial Regression for Infant Mortality Data) (Case Study 38 Regency/City in East Java)
Full text linkNegative binomial regression model is used to overcome the overdispersion in Poisson regression model. This model can be used to model the relationship of the infant mortality and the factors incidence. Geographical conditions, socio cultural and economic differ one of location another location causes the factors that influence infant mortality is different locally. Geographically Weighted Negative Binomial Regression (GWNBR) is one of methods for modeling that count data have spatial heterogeneity and overdispersion. The basic idea of this model considers of geography or location as the weight in parameter estimation. The parameter estimator is obtained from Iteratively Newton Raphson method. This research will determine the factors that influence infant mortality. GWNBR model with a weighting adaptive bi-square kernel function classifies regency/city in East Java into 16 groups based on the factors that significantly influence the number of infant mortality. This model is better used to analyze the number of infant mortality in East Java in 2008 due to a smallest deviance value.Keywords : Negative binomial regression, geographically weighted negative binomial regression, adaptive bi-square, overdispersio
PERFORMANCE COMPARISON BETWEEN KIMURA 2-PARAMETERS AND JUKES-CANTOR MODEL IN CONSTRUCTING PHYLOGENETIC TREE OF NEIGHBOUR JOINING
Full text linkBioinformatics as a recent improvement of knowledge has made an interest for scientist to collect and analyze data to provide the best estimate of the true phylogeny. The objective of this research is to construct and compare the phylogenetic tree of Neighbour Joining (NJ) based on different models (Kimura 2-Parameters and Jukes-Cantor) and to find out which model is more reliable on constructing NJ\u27s tree. In order to build the tree, reliable set of data is conducted from D-loop mtDNA sequences that is available in Gen Bank. The nucleotide sequences come from Bison bison (American bison), Bos taurus (European cow such as Shorthorn), Bos indicus (zebu breeds), Bos grunniens mutus (one of subspecies of cow), and Capra hircus (species of goat). The reliability of each models was measured using the Felsentein\u27s bootstrap method. The whole bootstrap process for each models was repeated 1.000, 5.000, and 10.000 times to detect its reliability. The performance was measured on the basis of the consistency of the topology relationship, the stability of nodes, the consistency of bootstrap confidence level (PB), standard error of distance, change of PB from (1.000-5.000) to (5.000-1.000), computational time, and BIC score. NJ\u27s phylogenetic tree with kimura 2-parameters and jukes cantor model have a good node stability and is also generally successful in representing topological relationships between taxa. The increasing of bootstrap replication number in common will increase the consistency of bootstrap confidence value ( . It means both models have a good reliability. But, when the number of sequences is large and the extent of sequence divergence is low, it is generally difficult to construct the tree by any models. In conclusion, Kimura 2-Parameters has a better performance than Jukes-Cantor. Key words: phylogenetic tree, Neighbour Joining, Kimura 2-Parameters, Jukes-Canto