Show simple item record

dc.contributor.advisorDjuraidah, Anik
dc.contributor.advisorFitrianto, Anwar
dc.contributor.authorYunita, Ria
dc.date.accessioned2026-07-12T13:34:00Z
dc.date.available2026-07-12T13:34:00Z
dc.date.issued2026
dc.identifier.urihttp://repository.ipb.ac.id/handle/123456789/174421
dc.description.abstractGeographically Weighted XGBoost (GWXGBoost) menggabungkan pembobotan spasial dengan kemampuan prediksi nonlinear XGBoost. Penelitian ini membandingkan kinerja GWXGBoost dengan fungsi objektif MSE dan Tweedie pada densitas ikan cakalang di Wilayah Pengelolaan Perikanan (WPP) 572–573. Tujuan penelitian adalah membandingkan kinerja kedua fungsi objektif, menentukan metode terbaik berdasarkan MAE dan RMSE, menentukan peubah penting tiap musim, serta menganalisis hasil prediksi densitas ikan cakalang. Data yang digunakan berupa grid spasial beresolusi 0,5°×0,5° dalam rentang Desember 2022–November 2023 yang diagregasi menjadi empat periode musim. Peubah respons yang digunakan untuk mewakili densitas ikan adalah catch per unit effort (CPUE), sedangkan peubah penjelas meliputi salinitas, suhu potensial laut, tinggi permukaan laut, arus utara–selatan dan timur–barat, oksigen terlarut, klorofil-a, produktivitas primer, dan kedalaman lapisan tercampur. Hasil penelitian menunjukkan bahwa fungsi objektif Tweedie menghasilkan MAE dan RMSE yang lebih rendah dibandingkan fungsi objektif MSE serta lebih mampu menangkap nilai ekstrem dan mengatasi autokorelasi spasial. Metode terbaik adalah GWXGBoost ensemble Tweedie dengan nilai MAE dan RMSE terendah disetiap periode. Prediksi densitas tinggi terkonsentrasi di WPP 572, terutama perairan barat Sumatra. Peubah penting global dan lokal bervariasi antarmusim, dengan produktivitas primer, suhu potensial laut, tinggi permukaan laut, salinitas, dan kedalaman lapisan tercampur menjadi faktor utama.
dc.description.abstractGeographically Weighted XGBoost (GWXGBoost) combines spatial weighting with the nonlinear predictive power of XGBoost. This study compared GWXGBoost performance using MSE and Tweedie objective functions for modeling skipjack tuna density in Fisheries Management Areas (FMA) 572–573. The objectives were to compare both objective functions, determine the best method based on MAE and RMSE, identify important variables per season, and analyze spatial prediction patterns of skipjack tuna density. Data comprised 0.5°×0.5° spatial grids for December 2022–November 2023, aggregated into four seasonal periods. Catch per unit effort (CPUE) served as the response variable, while explanatory variables included salinity, sea potential temperature, sea surface height, north–south and east–west currents, dissolved oxygen, chlorophyll-a, primary productivity, and mixed layer depth. Results showed that Tweedie produced lower MAE and RMSE than MSE, better capturing extreme values and addressing spatial autocorrelation. The best method was GWXGBoost ensemble Tweedie, with the lowest MAE and RMSE in every period. High-density predictions were concentrated in FMA 572, particularly off western Sumatra. Important global and local variables varied by season, with primary productivity, sea potential temperature, sea surface height, salinity, and mixed layer depth as dominant factors.
dc.description.sponsorship
dc.language.isoid
dc.publisherIPB Universityid
dc.titleEvaluasi Kinerja Geographically Weighted XGBoost Berbasis Fungsi Objektif Tweedie Densitas Ikan Cakalang di WPP 572-573id
dc.title.alternativeEvaluating Geographically Weighted XGBoost with a Tweedie Objective Skipjack Tuna Density in FMA 572–573
dc.typeSkripsi
dc.subject.keywordcatch per unit effortid
dc.subject.keywordgeographically weighted XGBoostid
dc.subject.keywordikan cakalangid
dc.subject.keywordfungsi objektif Tweedieid
dc.subject.keywordWPP 572-573id
dc.subtypeUndergraduate Theses


Files in this item

Thumbnail
Thumbnail
Thumbnail

This item appears in the following Collection(s)

Show simple item record