Sistem Prediksi Harga Crude Palm Oil (CPO) Menggunakan Arsitektur Multi-Layer Perceptron (MLP) dan IndoBERT Berbasis Data Web Scraping - Submit Jurnal/Seminar

Aryanditho, Rakasiwi (2026) Sistem Prediksi Harga Crude Palm Oil (CPO) Menggunakan Arsitektur Multi-Layer Perceptron (MLP) dan IndoBERT Berbasis Data Web Scraping - Submit Jurnal/Seminar. Bachelor thesis, Institut Teknologi Kalimantan.

	Text 04221073_Cover.pdf Restricted to Repository staff only until 4 October 2028. Download (469kB) | Request a copy
	Text 04221073_statement_of_authenticity.pdf Restricted to Repository staff only until 4 October 2028. Download (425kB) | Request a copy
	Text 04221073_publishing_agreement.pdf Restricted to Repository staff only until 4 October 2028. Download (368kB) | Request a copy
	Text 04221073_approval_sheet.pdf Restricted to Repository staff only until 4 October 2028. Download (1MB) | Request a copy
	Text 04221073_preface.pdf Restricted to Repository staff only until 4 October 2028. Download (718kB) | Request a copy
	Text 04221073_abstract_id.pdf Restricted to Repository staff only until 4 October 2028. Download (429kB) | Request a copy
	Text 04221073_abstract_en.pdf Restricted to Repository staff only until 4 October 2028. Download (398kB) | Request a copy
	Text 04221073_table_of_content.pdf Restricted to Repository staff only until 4 October 2028. Download (1MB) | Request a copy
	Text 04221073_illustrations.pdf Restricted to Repository staff only until 4 October 2028. Download (826kB) | Request a copy
	Text 04221073_tables.pdf Restricted to Repository staff only until 4 October 2028. Download (402kB) | Request a copy
	Text 04221073_notations.pdf Restricted to Repository staff only until 4 October 2028. Download (474kB) | Request a copy
	Text 04221073_chapter_1.pdf Restricted to Repository staff only until 4 October 2028. Download (1MB) | Request a copy
	Text 04221073_chapter_2.pdf Restricted to Repository staff only until 4 October 2028. Download (3MB) | Request a copy
	Text 04221073_chapter_3.pdf Restricted to Repository staff only until 4 October 2028. Download (2MB) | Request a copy
	Text 04221073_chapter_4.pdf Restricted to Repository staff only until 4 October 2028. Download (5MB) | Request a copy
	Text 04221073_conclusions.pdf Restricted to Repository staff only until 4 October 2028. Download (675kB) | Request a copy
	Text 04221073_bibliography.pdf Restricted to Repository staff only until 4 October 2028. Download (980kB) | Request a copy
	Text 04221073_enclosure.pdf Restricted to Repository staff only until 4 October 2028. Download (7MB) | Request a copy
	Text 04221073_paper.pdf Restricted to Repository staff only until 4 October 2028. Download (321kB) | Request a copy
	Text 04221073_presentation.pdf Restricted to Repository staff only until 4 October 2028. Download (4MB) | Request a copy
	Text 04221073_Form. TA-020.pdf Restricted to Repository staff only until 4 October 2028. Download (489kB) | Request a copy

Abstract

Indonesia merupakan produsen dan eksportir Crude Palm Oil (CPO) terbesar di dunia dengan kontribusi 52% dari total ekspor CPO global, sehingga stabilitas harganya menjadi faktor krusial bagi perekonomian nasional. Metode prediksi terdahulu seperti SVR dan Neural Network memiliki keterbatasan karena hanya mengandalkan data historis numerik tanpa mempertimbangkan faktor eksternal seperti sentimen berita pasar. Penelitian ini mengusulkan sistem prediksi harga CPO yang mengintegrasikan web scraping, analisis sentimen berbasis IndoBERT, dan arsitektur Multi-Layer Perceptron (MLP). Data dikumpulkan secara otomatis menggunakan Selenium dari MPOB.gov.my (856 baris data harga harian) dan Infosawit.com (3.764 baris headline berita), mencakup periode Januari 2024 hingga Mei 2026. Headline berita dilabeli ke dalam tiga kelas sentimen menggunakan LLaMA 3.1 via API Groq untuk men-fine-tune IndoBERT sebagai pengklasifikasi sentimen, dengan hasil akurasi 90,18% dan F1-Score 88,25%. Polarity score dari IndoBERT kemudian digabungkan dengan fitur teknikal numerik (harga historis, SMA, RSI) sebagai masukan model MLP guna memprediksi harga penutupan CPO satu hari ke depan. Evaluasi pada periode Mei 2026 menunjukkan model MLP+IndoBERT menghasilkan MAPE 0,76%, MAE Rp 149.467,12, dan RMSE Rp 192.191,60, unggul signifikan dibandingkan model MLP tanpa sentimen yang menghasilkan MAPE 1,33%, MAE Rp 260.936,99, dan RMSE Rp 315.801,12. Hasil ini membuktikan bahwa integrasi sentimen berita berbahasa Indonesia mampu meningkatkan akurasi prediksi harga CPO secara bermakna, khususnya dalam mendeteksi pembalikan arah harga pada kondisi pasar yang tidak stabil.

Item Type:	Thesis (Bachelor)
Subjects:	Q Science > QA Mathematics > QA76 Computer software T Technology > TK Electrical engineering. Electronics Nuclear engineering
Divisions:	Jurusan Teknologi Industri dan Proses > Teknik Elektro
Depositing User:	Aryanditho Rakasiwi
Date Deposited:	16 Jul 2026 03:19
Last Modified:	16 Jul 2026 03:19
URI:	http://repository.itk.ac.id/id/eprint/26593

Actions (login required)

View Item