Metode Regularisasi Untuk Spatial Poisson Point Process (Studi Kasus: Pemodelan Distribusi Spasial Pohon Beilschmiedia pendula Lauraceae di Pulau Barro Colorado)

Prabowo, Sigit Dwi (2021) Metode Regularisasi Untuk Spatial Poisson Point Process (Studi Kasus: Pemodelan Distribusi Spasial Pohon Beilschmiedia pendula Lauraceae di Pulau Barro Colorado). Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 06211850012009-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2023. Download (1MB) | Request a copy

Abstract

Data pola titik spasial terdapat dalam banyak konteks di mana fokus penelitiannya terletak dalam mendeskripsikan distribusi suatu event di ruang pengamatan, seperti lokasi keberadaan spesies pohon Beilschmiedia pendula Lauraceae di hutan. Penelitian ini berfokus pada estimasi paramater covariate pemodelan intensitas distribusi spesies atau disebut sebagai species distribution modeling (SDM) pada spesies pohon Beilschmiedia pendula Lauraceae di pulau Barro Corolado, dimana covariate dalam pemodelan ini berupa faktor lingkungan seperti variabel topologi dan kandungan sari tanah. Model inhomogeneous Poisson point process akan digunakan sebagai model spasial titik pohon Beilschmiedia pendula Lauracea. Estimasi parameter pada model ini mengunakan metode regularized maximum likelihood estimation yang dapat menghasilkan estimasi yang lebih optimal dibandingkan metode maximum likelihood estimation. Pemilihan model terbaik pada penelitian ini didapat dengan melakukan tuning parameter pertama dan kedua pada metode regularisasi. Kriteria tuning parameter menggunakan Bayesian Information Criterion (BIC) dengan nilai BIC terkecil digunakan sebagai kriteria model terbaik. Hasil analisis yang didapatkan bahwa metode regularized maximum likelihood estimation penalti adaptif elastic net dengan tuning parameter dapat digunakan dalam memilih kovariat dan mengestimasi parameter model. Model yang diperoleh pada kasus intensitas Pohon Beilschmiedia pendula Lauraceae memilih 59 kovariat dari 94 kovariat faktor lingkungan yang dianggap berpengaruh terhadap keberadaaan pohon Beilschmiedia pendula Lauraceae dengan nilai BIC model sebesar 40018,3.
====================================================================================================
Data on spatial point patterns exist in many contexts where the focus of the research lies in describing the distribution of an event in the observation room, such as the location of the species Beilschmiedia pendula Lauraceae in the forest. This study focuses on estimating covariate parameters for modeling the intensity of species distribution or known as species distribution modeling (SDM) in the Beilschmiedia pendula Lauraceae tree species on the Barro Corolado island, where the covariates in this modeling are environmental factors such as topological variables and soil extract content. The inhomogeneous Poisson point process model will be used as a spatial point model of the Beilschmiedia pendula Lauracea tree. The parameter estimation in this model uses the regularized maximum likelihood estimation method which can produce a more optimal estimate than the maximum likelihood estimation method. The selection of the best model in this study was obtained by tuning the first and second parameters of the regularization method. The tuning parameter criteria used the Bayesian Information Criterion (BIC) with the smallest BIC value used as the best model criteria. The results of the analysis show that the regularized maximum likelihood estimation method of adaptive elastic net penalty with tuning parameters can be used in selecting covariates and estimating model parameters. The model obtained in the case of Beilschmiedia pendula Lauraceae tree intensity selected 59 covariates from 94 covariates of environmental factors that were considered influencing the existence of Beilschmiedia pendula Lauraceae trees with a BIC model value of 40018.3.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Bayesian Information Criterion (BIC), regularized maximum likelihood estimation, species distribution modeling, Bayesian Information Criterion (BIC), regularized maximum likelihood estimation, species distribution modeling.
Subjects:	H Social Sciences > HA Statistics > HA30.6 Spatial analysis
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49101-(S2) Master Thesis
Depositing User:	Sigit Dwi Prabowo
Date Deposited:	13 Mar 2021 02:13
Last Modified:	13 Mar 2021 02:13
URI:	http://repository.its.ac.id/id/eprint/84161

Actions (login required)

View Item