<![CDATA[This study aims to analyze and compare the performance of the K-Means and K-Medoids algorithms in clustering Cookpad application user reviews obtained directly from the Google Play Store. A total of 15,330 reviews were collected as a representation of authentic user opinions. The review text was first processed through a pre-processing stage, then converted into a numeric representation using the TF-IDF method. The clustering process was carried out using the K-Means and K-Medoids algorithms, which were then evaluated using the Silhouette Score metric, both on the original data and after removing outliers with the DBSCAN algorithm. The results showed that K-Medoids performed better with a score of 0.2788 (with outliers) and 0.2966 (without outliers), compared to K-Means which obtained a score of 0.2625 (with outliers) and 0.2795 (without outliers). The clusters formed were dominated by positive reviews with words such as "cook", "help", and "recipe", which describe user satisfaction. This study provides a systematic approach to analyzing Google Play app reviews by directly comparing two clustering algorithms and integrating outlier evaluation to improve the quality of cluster results. The limitation is that the data only comes from one platform. The results are visualized through an interactive web interface and are expected to be input in developing application features based on user needs.]]>
Copyrights © 2025
