Pemetaan Sentimen Media dan Hubungan Entitas dalam Berita Teknologi Pertanian

Mufida, Elisa Nur Syafiatul (2025) Pemetaan Sentimen Media dan Hubungan Entitas dalam Berita Teknologi Pertanian. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6026232010-Master_Tesis.pdf - Draft Version Download (17MB)

Abstract

Digitalisasi pertanian memiliki perkembangan yang begitu pesat dengan adanya dukungan teknologi seperti Big Data, Internet of Things, dan Artificial Intelligence yang didorong oleh peran media sebagai pembentuk opini publik. Penelitian ini mengintegrasikan Machine Learning, dan Knowledge Graph untuk memetakan sentimen dan relasi antar entitas dalam 1.295 artikel oleh BBC, The Guardian, The Independent, dan Reuters, pada publikasi 2013-2025. Hasil klasifikasi menunjukan model BERT dengan Random Forest unggul dengan hasil akurasi 92% dan F1-macro rata-rata 87.2% menjadi baseline terbaik dibanding dengan BERT + SVM (91%), dan BERT+XGBoost (91%). Pada tahap pemetaan relasi, triples hasil NER dan Relation Extraction dibangun pada Knowledge Graph, kemudian dieksplorasi menggunakan Graph Convolutional Network. Klaster K-Means menunjukkan lebih dari 70% relasi masih berupa open triad, menandakan celah koneksi. Enrichment diikuti pemisahan komunitas Louvain meningkatkan modularitas menjadi 0,45 dan membentuk empat komunitas yang sejalan dengan enam topik utama: Food Security & Risk Management, Economic & Family Farming, Supply Chain & Trade, Technology Adoption, Climate & Environment, serta Investment & Market. Sentimen Pro mendominasi 70% pada topik ketahanan pangan, sedangkan komunitas perdagangan memuat porsi Contra tertinggi, yakni 40%. Secara keseluruhan, pendekatan gabungan KG–GCN–ML efektif memperkaya analisis sentimen berita teknologi pertanian, mengungkap pola topik sentimen sekaligus peluang perluasan relasi antar entitas untuk riset kebijakan dan strategi adopsi teknologi pertanian.
=====================================================================================================================================
Agricultural digitalization propelled by IoT, AI, and Big Data, and amplified by media influence is rapidly evolving. This study combines Knowledge Graphs (KG) and Machine Learning to map sentiment across 1,295 articles news from BBC, The Guardian, The Independent, and Reuters published between 2013 and 2025. The best classifier, a BERT model paired with a Random Forest, attains 92% accuracy and an 88 %F1-macro, outperforming BERT-SVM and BERT-XGBoost. For relation mapping, NER- and RE-derived triples are built into a KG and explored with a Graph Convolutional Network. Initial k-means clustering shows that over 70 % of the triads are open, revealing connectivity gaps; community enrichment with Louvain raises modularity to 0.45 and yields four communities aligned with six key topics Food Security and Risk Management; Economic and Family Farming; Supply Chain and Trade; Technology Adoption; Climate and Environment; and Investment and Market. Pro sentiment dominates food-security coverage at roughly 70 %, while trade-focused articles carry the highest Contra share, at 40 %. Overall, the integrated KG–GCN–ML framework enriches agritech sentiment analysis, unveiling topic–sentiment patterns and opportunities for expanding entity relations in policy and technology-adoption research.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Bidirectional Encorder Representations from Transformers, RandomForest, Konowledge Graph, XGBoost, Support Vector Machine.
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 59101-(S2) Master Thesis
Depositing User:	Elisa Nur Syafiatul Mufida
Date Deposited:	06 Aug 2025 06:37
Last Modified:	06 Aug 2025 06:37
URI:	http://repository.its.ac.id/id/eprint/127751

Actions (login required)

View Item