Please use this identifier to cite or link to this item:
http://repository.ipb.ac.id/handle/123456789/163634Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor.advisor | Seminar, Kudang Boro | |
| dc.contributor.advisor | Sudradjat | |
| dc.contributor.author | Hakim, Muhammad Wildan Bagir | |
| dc.date.accessioned | 2025-07-02T14:14:51Z | |
| dc.date.available | 2025-07-02T14:14:51Z | |
| dc.date.issued | 2025 | |
| dc.identifier.uri | http://repository.ipb.ac.id/handle/123456789/163634 | |
| dc.description.abstract | Pemantauan status hara seng (Zn) pada tanaman kelapa sawit merupakan aspek krusial dalam menunjang produktivitas dan efisiensi pemupukan. Namun, metode konvensional saat ini bersifat destruktif sehingga tidak efisien untuk diterapkan dalam skala luas. Penelitian ini bertujuan mengembangkan model prediksi Zn berbasis algoritma cerdas Extreme Gradient Boosting (XGBoost) dan Deep Neural Network (DNN) guna menentukan status Zn pada kelapa sawit di lahan mineral, gambut, serta kombinasi keduanya. Data yang digunakan mencakup citra Sentinel-1A (radar), Sentinel-2A (optik), serta indeks vegetasi sebagai variabel bebas, dan kandungan Zn pada daun hasil analisis laboratorium sebagai variabel terikat. Hasil penelitian menunjukkan bahwa model XGBoost secara konsisten memberikan performa yang lebih unggul dibandingkan DNN. Pada lahan mineral, XGBoost mencapai akurasi (correctness) sebesar 91,35%, sedangkan DNN hanya 88,23%. Pada lahan gambut, XGBoost mencatatkan akurasi 87,95%, lebih tinggi dibandingkan DNN sebesar 83,46%. Sementara itu, pada model gabungan, XGBoost tetap unggul dengan akurasi 87,21% dibandingkan DNN yang memperoleh 85,97%. Pengujian model gabungan terhadap data spesifik lahan menunjukkan bahwa XGBoost mampu mempertahankan akurasi tinggi, yaitu 90,09% untuk lahan mineral dan 86,36% untuk lahan gambut, dengan selisih performa yang kecil dibandingkan DNN. Model gabungan dapat menjadi alternatif prediksi praktis yang layak meski dengan penurunan correctness sebesar 1,2-1,5%. Secara keseluruhan, model yang dikembangkan mampu memetakan variasi spasial kandungan Zn dan merepresentasikan perbedaan karakteristik antara lahan mineral dan gambut. Hasil ini dapat memberikan gambaran dan menjadi acuan dalam manajemen pemupukan di perkebunan kelapa sawit yang lebih presisi. | |
| dc.description.abstract | Monitoring the zinc (Zn) nutrient status in oil palm is a crucial aspect of supporting productivity and fertilization efficiency. However, conventional monitoring methods are currently destructive and therefore inefficient for large scale application. This study aims to develop Zn prediction models based on smart algorithms, namely Extreme Gradient Boosting (XGBoost) and Deep Neural Network (DNN), to determine Zn status in oil palm cultivated on mineral soil, peatland, and a combination of both. The input data include Sentinel-1A (radar), Sentinel-2A (optical), and vegetation indices as independent variables, while Zn concentrations in palm leaves obtained through laboratory analysis serve as the dependent variable. The results show that the XGBoost model consistently outperforms the DNN model. On mineral soil, XGBoost achieved an accuracy (correctness) of 91,35%, whereas DNN reached only 88,23%. On peatland, XGBoost recorded an accuracy of 87,95%, surpassing DNN at 83,46%. In the combined model, XGBoost remained superior with an accuracy of 87,21%, compared to DNN's 85,97%. Further testing of the combined model using specific land data showed that XGBoost maintained high accuracy of 90,09% for mineral soil and 86,36% for peatland with only slight performance differences compared to DNN. Despite a 1,2–1,5% decrease in accuracy, the combined model remains a viable and practical alternative for Zn prediction across diverse land types. Overall, the developed model successfully maps the spatial variability of Zn content and reflects the distinct characteristics between mineral and peat soils. These results can provide insights and serve as a reference for more precise fertilization management in oil palm plantations. | |
| dc.description.sponsorship | Tim Riset PreciPalm (Precision Agriculture Platform for Oil Palm) | |
| dc.language.iso | id | |
| dc.publisher | IPB University | id |
| dc.title | Sinergis Sentinel-1A dan Sentinel-2A untuk Penentuan Status Hara Seng (Zn) Daun Kelapa Sawit dengan Algoritma Sistem Cerdas | id |
| dc.title.alternative | Synergise of Sentinel-1A and Sentinel-2A for Determine Zinc (Zn) Nutrient Status in Oil Palm Leaf Using Smart System Algorithm | |
| dc.type | Skripsi | |
| dc.subject.keyword | oil palm | id |
| dc.subject.keyword | Sentinel | id |
| dc.subject.keyword | zinc | id |
| dc.subject.keyword | deep neural network | id |
| dc.subject.keyword | extreme gradient boosting | id |
| Appears in Collections: | UT - Agricultural and Biosystem Engineering | |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| cover_F1401211044_aeac07ed4ab84919a631807312153f81.pdf | Cover | 552.22 kB | Adobe PDF | View/Open |
| fulltext_F1401211044_4331da7713c34861979cdb68966c191e.pdf Restricted Access | Fulltext | 3.11 MB | Adobe PDF | View/Open |
| lampiran_F1401211044_44c39abaaa664e92a1fb573e36e11069.pdf Restricted Access | Lampiran | 1.74 MB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.