Analysis of Deep K-Means and K-Means for Journal Summarization Using the BERTopic Method

Abstract

This research aims to analyze the performance of Deep K-means and K-means methods in summarizing and clustering final project abstracts using the BERTopic technique. The methods used include preprocessing abstract data, applying Deep K-means and K-means clustering algorithms, and modeling topics with BERTopic. The dataset used was obtained from Kaggle and analyzed using the Python programming language. The results show that the Deep K-means method is able to produce more coherent topic clusters than conventional K-Means, with higher Silhouette Score values. The combination of clustering and topic modeling techniques proved effective in automatically summarizing and grouping final project abstracts, making it easier to identify student research trends. This study concludes that a hybrid approach using Deep K-Means, K-Means, and BERTopic can be a promising solution for the thematic analysis of final projects in academic environments. The combination of Deep K-means with BERTopic emerges as a highly promising solution for automated thematic analysis in academic environments, paving the way for more efficient research workflows.