Smart Learning Library

  • Beranda
  • Informasi
  • News
  • Bantuan
  • Pustakawan
  • Area Anggota
  • Pilih Bahasa :
    Arabic Bengali Brazilian Portuguese English Espanol German Indonesia Bahasa Jepang Melayu Persia Russian Thai Turkish Urdu

Search by:

All Author Subject ISBN/ISSN Advanced Search

Last search:

{{tmpObj[k].text}}
Image of EVALUASI METODE PERMUTATION FEATURE IMPORTANCE PADA DETEKSI PENYAKIT TANAMAN

SKRIPSI IF

EVALUASI METODE PERMUTATION FEATURE IMPORTANCE PADA DETEKSI PENYAKIT TANAMAN

ASHARI, MUWAHID ZAKI - Personal Name;

ABSTRAK
Perkembangan model machine learning untuk deteksi penyakit tanaman kerap menghadapi permasalahan trade-off antara tingkat akurasi dan interpretabilitas, di mana model yang lebih akurat cenderung sulit dijelaskan dasar pengambilan keputusannya. Penelitian ini bertujuan untuk mengevaluasi trade-off tersebut dengan membandingkan model Random Forest dan Neural Network pada klasifikasi penyakit tanaman kentang dan tomat menggunakan metode Permutation Feature Importance (PFI) sebagai instrumen interpretasi yang bersifat model-agnostic. Penelitian menggunakan dataset PlantVillage sebanyak 18.163 gambar daun yang tersebar pada 13 kelas, dengan sepuluh fitur hand-crafted yang mencakup karakteristik warna, tekstur, dan rasio area daun sebagai representasi masukan bagi kedua model. Hasil penelitian menunjukkan bahwa Neural Network memperoleh akurasi uji sebesar 92,4% dan rata-rata akurasi cross-validation 5-fold sebesar 90,3%, unggul dibandingkan Random Forest yang memperoleh 82,2% dan 82,3%. Namun, hasil PFI menunjukkan tingkat konsistensi fitur terpenting yang rendah antara kedua model, di mana Random Forest secara konsisten mengandalkan fitur tekstur dan rasio area warna yang mudah dijelaskan secara biologis, sedangkan Neural Network lebih mengandalkan intensitas warna secara global. Penelitian ini menyimpulkan bahwa peningkatan akurasi pada model yang lebih kompleks tidak sebanding dengan penurunan interpretabilitas yang terjadi, sehingga pemilihan model perlu disesuaikan dengan kebutuhan pengguna, baik untuk kebutuhan akurasi maupun kebutuhan penjelasan yang dapat dipertanggungjawabkan secara biologis.
Kata Kunci: Permutation Feature Importance, Random Forest, Neural Network, interpretabilitas, penyakit tanaman, trade-off.

ABSTRACT
The development of machine learning models for plant disease detection often faces a trade-off between accuracy and interpretability, where more accurate models tend to be difficult to explain. This study aims to evaluate this trade-off by comparing Random Forest and Neural Network models on the classification of potato and tomato plant diseases using the Permutation Feature Importance (PFI) method as a model-agnostic interpretation instrument. The study utilized the PlantVillage dataset consisting of 18,163 leaf images spread across 13 classes, with ten hand-crafted features covering color, texture, and leaf area ratio characteristics as input representations for both models. The results show that Neural Network achieved a test accuracy of 92.4% and a mean 5-fold cross-validation accuracy of 90.3%, outperforming Random Forest, which achieved 82.2% and 82.3% respectively. However, the PFI results show a low level of consistency in the most important features between the two models, where Random Forest consistently relies on texture and color area ratio features that are biologically explainable, while Neural Network relies more on overall color intensity. This study concludes that the accuracy improvement obtained by the more complex model is not proportional to the resulting decrease in interpretability, so that model selection should be tailored to user needs, whether the priority is predictive accuracy or biologically accountable explanations.
Keywords: Permutation Feature Importance, Random Forest, Neural Network, interpretability, plant disease, trade-off.


Ketersediaan
S260908018607.2 ASH ePerpustakaan STMIK AMIKBANDUNGTersedia
Informasi Detil
Judul Seri
-
No. Panggil
607.2 ASH e
Penerbit
: ., 2026
Deskripsi Fisik
-
Bahasa
Indonesia
ISBN/ISSN
-
Klasifikasi
NONE
Tipe Isi
-
Tipe Media
-
Tipe Pembawa
-
Edisi
-
Subyek
SKRIPSI IF
Info Detil Spesifik
-
Pernyataan Tanggungjawab
-
Versi lain/terkait

Tidak tersedia versi lain

Lampiran Berkas
Komentar

You must be logged in to post a comment

Smart Learning Library
  • Information
  • Services
  • Librarian
  • Member Area

About Us

As a complete Library Management System, SLiMS (Senayan Library Management System) has many features that will help libraries and librarians to do their job easily and quickly. Follow this link to show some features provided by SLiMS.

Search

start it by typing one or more keywords for title, author or subject

Keep SLiMS Alive Want to Contribute?

© 2026 — Senayan Developer Community

Powered by SLiMS
Select the topic you are interested in
  • Computer science, information & general works
  • Philosophy & psychology
  • Religion
  • Social sciences
  • Language
  • Pure Science
  • Applied sciences
  • Arts & recreation
  • Literature
  • History & geography
Advanced Search