Show simple item record

dc.contributor.advisorMasjkur, Mohammad
dc.contributor.advisorSartono, Bagus
dc.contributor.authorAisyah, Usthuanah
dc.date.accessioned2023-08-01T15:47:02Z
dc.date.available2023-08-01T15:47:02Z
dc.date.issued2023
dc.identifier.urihttp://repository.ipb.ac.id/handle/123456789/122947
dc.description.abstractSeleksi peubah berguna untuk memilih subset peubah yang revelan terhadap peubah respon, dan seleksi peubah dapat diterapkan untuk mengetahui peringkat peubah yang berhubungan erat dengan suatu kejadian, salah satunya rawan pangan. Penelitian ini bertujuan menyaring peubah yang berhubungan erat dengan kejadian rawan pangan di Jawa Barat berdasarkan algoritme filter univariat information gain, gain ratio, symmetrical uncertainty, gini gain, dan chi-square serta mengidentifikasi subset peubah terbaik berdasarkan model klasifikasi yang menghasilkan ukuran kinerja paling baik menggunakan algoritme klasifikasi Xtreme Gradient Boosting (XGBoost). Hasil penelitian menunjukkan status kepemilikan kulkas, komputer, emas minimal 10 gram, dan televisi layar datar minimal 30 inch merupakan peubah yang konsisten terpilih pada semua algoritme filter univariat tersebut. Pemodelan klasifikasi XGBoost dengan subset peubah terpilih gain ratio menghasilkan nilai sensitivitas dan spesifisitas yang seimbang sebesar 61,9% dan 61,4%. Hasil penyaringan peubah gain ratio mengindikasikan bahwa rumah tangga yang tidak memiliki modal ekonomi berupa emas minimal 10 gram, komputer, kulkas, televisi layar datar minimal 30 inch dan rumah tangga penerima Bantuan Pangan Nontunai (BPNT) cenderung rawan pangan.id
dc.description.abstractFeature selection is useful for selecting a subset of variables relevant to the response variable, and it can be applied to determine the rank of variables closely related to an incident, one of which is food insecurity. This study aims to select variables closely related to food insecurity in West Java based on information gain, gain ration, symmetrical uncertainty, gini gain, and chi-square univariate filter algorithms and identify the best subset of variables based on the classification model that produces the best performance measures using the Xtreme Gradient Boosting (XGBoost) algorithm. The study result shows that the ownership status of a refrigerator, a computer, a gold of at least 10 grams, and a flat-screen television of at least 30 inches are the consistent variables selected on all the univariate filter algorithms. XGBoost classification modeling with the subset of selected variables gain ratio has balanced sensitivity and specificity values of 61,9% and 61,4%. The gain ratio variables filtering result indicates that households that do not have economic capital in the form of gold at least 10 grams, computer, refrigerator, flat-screen television at least 30 inches, and households receiving Non-Cash Food Aid (BPNT) tend to be food insecure.id
dc.description.sponsorshipKementerian Pendidikan, Kebudayaan, Riset dan Teknologiid
dc.language.isoidid
dc.publisherIPB Universityid
dc.titlePenerapan Metode Filter Feature Selection untuk Menentukan Subset Peubah Terbaik dalam Pemodelan Kejadian Rawan Pangan di Jawa Baratid
dc.title.alternativeThe Implementation of Filter Feature Selection Method to Identify the Best Subset of Variables in Modeling of Food Insecurity in West Javaid
dc.typeUndergraduate Thesisid
dc.subject.keywordclassificationid
dc.subject.keywordfeature selectionid
dc.subject.keywordfood insecurityid
dc.subject.keywordunivariate methodid
dc.subject.keywordxgboostid


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record