IKN News Topic Analysis of Nusantara Capital City using Frobenius Norm and Non-negative Matrix Factorization

Abstract

This research leverages Non-negative Matrix Factorization (NMF) with the Frobenius norm to analyze news articles from Kompas about the relocation of Indonesia's capital to Nusantara. The study is significant as it provides insights into public and media perceptions documented by Kompas, identifies critical issues surrounding this transformative national project, and demonstrates the utility of NMF in analyzing Indonesian-language news texts, particularly in the context of public policy and media discourse. A dataset of news articles related to Ibu Kota Nusantara was preprocessed through cleaning, normalization, stemming/lemmatization, and tokenization to prepare it for topic modeling. Using TF-IDF for feature extraction, Non-Negative Matrix Factorization (NMF) with Frobenius norm as the loss function was applied to generate topics, which were evaluated based on coherence scores and manual analysis for relevance and interpretability. This study identified five distinct topics related to Ibu Kota Nusantara (IKN) from Kompas news articles during January-March 2024, covering community preparations, toll road developments, buffer zone status, groundbreaking events, and ASN housing. Using the NMF model and c_uci coherence scoring, the study achieved a high coherence score of 0.991, indicating semantically connected terms that facilitate topic interpretation. The alignment between Wordcloud and NMF results demonstrates both methods' focus on significant terms, with Wordcloud highlighting key words and NMF providing a deeper structural analysis of topic interrelations.