Penerapan LSI untuk Topic Modelling Skripsi Sistem Informasi di Beberapa PTN Surabaya

Abstract

This study applies Latent Semantic Indexing (LSI) for topic modelling on undergraduate thesis titles from Information Systems students at four public universities in Surabaya. The purpose of this research is to explore dominant research themes and academic trends that have emerged over recent years. A text mining approach is employed, where hundreds of thesis titles are collected from each university and pre-processed through tokenization, stopword removal, stemming, and term weighting using TF-IDF. The LSI method is then used to extract latent topics by reducing the dimensionality of the term-document matrix through singular value decomposition (SVD). The results indicate the presence of several dominant topics such as information system development, decision support systems, and data mining. These topics reflect the recurring areas of interest in the Information Systems curriculum across the universities studied. The study concludes that LSI is effective in identifying hidden semantic patterns and grouping thesis topics into meaningful clusters. These findings may support curriculum development and academic planning by highlighting students' thematic focus and institutional research directions.