Please use this identifier to cite or link to this item:
http://repository.ipb.ac.id/handle/123456789/162457Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor.advisor | Erfiani | - |
| dc.contributor.advisor | Sumertajaya, I Made | - |
| dc.contributor.author | Suruddin, Adzkar Adlu Hasyr | - |
| dc.date.accessioned | 2025-06-12T08:28:46Z | - |
| dc.date.available | 2025-06-12T08:28:46Z | - |
| dc.date.issued | 2025 | - |
| dc.identifier.uri | http://repository.ipb.ac.id/handle/123456789/162457 | - |
| dc.description.abstract | Penelitian ini bertujuan untuk membandingkan kinerja metode RK LASSO dan RK SIR-LASSO dalam konteks pemodelan data berdimensi rendah hingga tinggi. Evaluasi dilakukan melalui simulasi serta penerapan pada data pengukuran kadar glukosa darah non-invasif tahun 2019. Hasil simulasi menunjukkan bahwa RK LASSO memiliki keunggulan dalam beberapa kondisi, terutama pada data berdimensi rendah. Namun, secara umum, RK SIR-LASSO menunjukkan performa yang lebih unggul, khususnya dalam kondisi jumlah peubah yang besar dan korelasi antar peubah yang tinggi. Kinerja kedua metode cenderung menurun ketika korelasi antar peubah (?) tinggi, yang mengindikasikan adanya multikolinearitas yang menyulitkan pemisahan pengaruh peubah terhadap respon. RK SIR-LASSO memberikan nilai Root Mean Square Error of Prediction (RMSEP) yang lebih rendah dan stabilitas prediksi yang lebih tinggi dibandingkan RK LASSO. Dalam penerapan pada data nyata, RK SIR-LASSO menunjukkan konsistensi terbaik pada jumlah slice optimal H = 14. Temuan ini memperkuat kajian teoritis mengenai pentingnya pemilihan jumlah slice yang tepat untuk menghasilkan prediksi yang optimal. Dengan demikian, RK SIR-LASSO direkomendasikan sebagai metode yang lebih andal dalam pemodelan data dengan kompleksitas tinggi. | - |
| dc.description.abstract | The aim of this research is to compare the performance of the CR LASSO) and CR SIR-LASSO) methods in modeling low to high-dimensional data. The comparison was carried out via simulations and using the application of non-invasive blood glucose level monitoring data from 2019. From simulation, CR LASSO is seen to perform better under some conditions, especially for low-dimensional data. But overall, CR SIR-LASSO performs better, particularly when the number of variables is large and there is high correlation among variables. Both the methods perform worse when there is high correlation among variables (?) or, equivalently, when there is multicollinearity among variables, and it is difficult to separate the effect of each variable on the response. CR SIR-LASSO has smaller Root Mean Square Error of Prediction (RMSEP) values and more stable predictions than CR LASSO. For real data application, CR SIR-LASSO is most stable at optimum slices number, H = 14. The findings confirm earlier theoretical research on the significance of the correct selection of slices number in order to reach optimal prediction. Hence, CR SIR-LASSO is suggested as a more stable approach to model complicated data. | - |
| dc.description.sponsorship | null | - |
| dc.language.iso | id | - |
| dc.publisher | IPB University | id |
| dc.title | Regresi Kontinum Menggunakan Prapemrosesan Seleksi Peubah LASSO dan SIR-LASSO pada Alat Keluaran Glukosa Darah Non-Invasif | id |
| dc.title.alternative | null | - |
| dc.type | Tesis | - |
| dc.subject.keyword | model kalibrasi | id |
| dc.subject.keyword | LASSO | id |
| dc.subject.keyword | multikolinearitas | id |
| dc.subject.keyword | sliced inverse regression | id |
| Appears in Collections: | MT - School of Data Science, Mathematic and Informatics | |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| cover_G1501211033_144561f84a794adbbbe6733ee687fbdf.pdf | Cover | 2.13 MB | Adobe PDF | View/Open |
| fulltext_G1501211033_5cabaa7eea74496881582a408e2f7382.pdf Restricted Access | Fulltext | 2.57 MB | Adobe PDF | View/Open |
| lampiran_G1501211033_630f61bd70d646dca7baa3255e430430.pdf Restricted Access | Lampiran | 1.58 MB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.