IMPLEMENTASI METODE LATENT DIRICLHET ALLOCATION (LDA) DALAM TOPIC MODELING PADA DATA BERITA KESEHATAN

The use of the internet and digital media has transformed news consumption from traditional sources to digital platforms like Facebook, Twitter, and Instagram, posing challenges related to the validity and reliability of information. News sites like kompas.com have adapted by offering multimedia formats and premium subscription services. In the context of health news, the complexity of topics and the variety of terms present challenges, necessitating detailed grouping. Latent Dirichlet Allocation (LDA) is an effective method for clustering health news and topic analysis. The research process involves problem identification, data collection, data preprocessing, LDA model creation, and evaluation. The LDA analysis of health news data identifies five main topics: Topic 0: body health (keywords: "signs", "alert", "attention", "body"). Topic 1: blood issues (keywords: "blood", "sugar"). Topic 2: disease symptoms (keywords: "know", "symptoms", "sick", "recover"). Topic 3: healthy drinks (keywords: "effects", "drink"). Topic 4: types of diseases (keywords: "blood", "sugar", "sleep", "eat", "fruit"). The implementation of LDA is conducted through a website using Streamlit, allowing for data upload, text preprocessing, and LDA application. A trial with new data addition and stemming removal shows that the analysis results are more accurate and relevant, retaining the original meaning of words and improving the quality of the analysis.
Keywords: Internet, Latent Dirichlet Allocation (LDA), topic modeling.