Show simple item record

dc.contributor.advisorDasanto, Bambang Dwi
dc.contributor.authorAFDAL, ADWIN ZAKARIA
dc.date.accessioned2026-06-26T03:37:02Z
dc.date.available2026-06-26T03:37:02Z
dc.date.issued2026
dc.identifier.urihttp://repository.ipb.ac.id/handle/123456789/173712
dc.description.abstractWilayah selatan Indonesia merupakan salah satu pusat formasi siklon tropis di Samudra Hindia dan dunia, namun belum terdapat model prediksi yang secara khusus dikembangkan untuk wilayah ini. Penelitian ini bertujuan untuk mengembangkan dan mengevaluasi model probabilistik untuk prediksi siklogenesis menggunakan regresi logistik multivariat, yang diintegrasikan dengan metode Mutual Information Feature Selection (MI-FS) dan algoritma Minimum Redundancy Maximum Relevance (mRMR), dengan berbagai variabel atmosfer, oseanik, dan variabilitas iklim skala besar sebagai prediktor. Hasil penelitian menunjukkan bahwa variabel dinamika atmosfer lapisan bawah, khususnya ketinggian geopotensial, kecepatan vertikal, vortisitas relatif, dan divergensi, mendominasi dalam prediksi formasi siklon tropis, diikuti oleh variabel dinamika atmosfer lapisan atas, termodinamika atmosfer, serta variabilitas iklim skala besar. Peningkatan jumlah prediktor meningkatkan kemampuan diskriminasi model, sementara kemampuan probabilistik relatif tidak berubah. Konfigurasi model optimal diperoleh pada penggunaan sembilan prediktor teratas (Model 9), yakni ketinggian geopotensial, kecepatan vertikal, RMM2, vortisitas relatif, divergensi, vortisitas potensial, RMM1, vertical wind shear, dan ENSO, dengan skor performa multi-metrik tertinggi sebesar 0,97994, karena mampu merepresentasikan faktor utama formasi siklon tropis tanpa redundansi informasi yang berlebihan.
dc.description.abstractThe southern region of Indonesia is one of the centers of tropical cyclone formation in the Indian Ocean and the world, however, there is no dedicated predictive model that has been developed specifically for this region. This study aims to develops and evaluates a probabilistic model for tropical cyclone genesis prediction using multivariate logistic regression, with the integration of Mutual Information Feature Selection (MI-FS) and Minimum Redundancy Maximum Relevance (mRMR) algorithm, with various atmospheric, oceanic, and large-scale climate variability as predictors. The result indicate that lower-atmospheric dynamical variables, particularly geopotential height, vertical velocity, relative vorticity, and divergence dominate the prediction of tropical cyclone formation, followed by upper-atmosphere dynamics, atmospheric thermodynamics, and large-scale climate variability variables. The increase of the number of predictors improve the model’s discrimination ability, while probabilistic skill remain unchanged. The optimal model configuration is obtained using the top nine (Model 9) predictors, namely geopotential height, vertical velocity, RMM2, relative vorticity, divergence, potential vorticity, RMM1, vertical wind shear, and ENSO, yielding the highest multi-metric performance score of 0,97994 by effectively representing the primary drivers of tropical cyclone genesis without excessive information redundancy.
dc.description.sponsorship
dc.language.isoid
dc.publisherIPB Universityid
dc.titlePengembangan Model Prediksi Siklon Tropis Berbasis Regresi Logistik yang Terintegrasi dengan MI-FS dan mRMRid
dc.title.alternativeDevelopment of a Tropical Cyclone Prediction Model Based on Logistic Regression Integrated with MI-FS and mRMR
dc.typeSkripsi
dc.subject.keywordsiklogenesisid
dc.subject.keywordregresi logistikid
dc.subject.keywordseleksi fiturid
dc.subject.keyworddinamika atmosferid
dc.subject.keywordvariabilitas iklimid


Files in this item

Thumbnail
Thumbnail
Thumbnail

This item appears in the following Collection(s)

Show simple item record