PENGELOMPOKAN ULASAN VIDEO GAME ELDEN RING PADA PLATFORM STEAM DENGAN METODE SPECTRAL CLUSTERING MENGGUNAKAN PRINCIPAL COMPONENT ANALYSIS BERBASIS SIMILARITY GAUSSIAN

CLUSTERING OF ELDEN RING VIDEO GAME REVIEWS ON THE STEAM PLATFORM USING SPECTRAL CLUSTERING METHOD WITH PRINCIPAL COMPONENT ANALYSIS BASED ON GAUSSIAN SIMILARITY

Perkembangan teknologi yang pesat dalam beberapa tahun terakhir telah mendorong kemajuan signifikan dalam industri video game. Video game yang sebelumnya hanya dapat dimainkan secara offline kini bisa diakses secara online berkat internet. Popularitas video game meningkat seiring dengan bertambahnya jumlah pemain, yang pada tahun 2020 mencapai 3,5 miliar orang, dengan mayoritas (69 persen) bermain di mobile. Pendistribusian video game juga beralih dari media fisik ke platform digital, seperti Steam, yang memungkinkan pemain membeli, memainkan, dan mengulas game. Fitur ulasan di Steam menjadi sarana penting bagi pengembang untuk meningkatkan kualitas game berdasarkan umpan balik pemain. Namun, jumlah ulasan yang sangat banyak dapat menyulitkan analisis secara manual. Oleh karena itu, diperlukan pengelompokan ulasan agar ulasan dengan tema serupa dapat dikelompokkan, memudahkan pengembang dalam memahami dan merespons feedback. Penelitian ini mengaplikasikan metode Spectral Clustering yang dikombinasikan dengan Principal Component Analysis (PCA) berbasis Similarity Gaussian untuk mengelompokkan ulasan video game Elden Ring di Steam. Metode ini dipilih karena kemampuannya menangani data kompleks dan besar, serta menghasilkan pengelompokan yang lebih baik. Evaluasi performa klaster dilakukan menggunakan Silhouette Coefficient, menghasilkan skor terbaik 0,5077369613468565 dengan tiga klaster utama: Klaster 0 (ulasan sangat positif), Klaster 1 (ulasan beragam aspek game), dan Klaster 2 (ulasan aspek gameplay dan teknis). Hasil penelitian menunjukkan bahwa metode Spectral Clustering dengan PCA berbasis Similarity Gaussian efektif dalam mengelompokkan ulasan video game, sehingga dapat membantu pengembang meningkatkan kualitas game berdasarkan pola yang teridentifikasi dari klaster ulasan. 

The rapid technological advancements in recent years have significantly propelled the video game industry forward. Video games that were once only playable offline can now be accessed online thanks to the internet. The popularity of video games has increased alongside the growing number of players, which reached 3.5 billion people in 2020, with the majority (69 percent) playing on mobile devices. Game distribution has also shifted from physical media to digital platforms, such as Steam, allowing players to buy, play, and review games. The review feature on Steam is a crucial tool for developers to improve game quality based on player feedback. However, the vast number of reviews makes manual analysis difficult. Therefore, it is necessary to group reviews so that those with similar themes can be clustered together, making it easier for developers to understand and respond to feedback. This study applies the Spectral Clustering method combined with Principal Component Analysis (PCA) based on Gaussian Similarity to cluster reviews of the video game Elden Ring on Steam. This method was chosen for its ability to handle complex and large data sets, as well as its superior clustering performance. The clustering performance was evaluated using the Silhouette Coefficient, yielding the best score of 0.5077369613468565 with three main clusters: Cluster 0 (highly positive reviews), Cluster 1 (reviews on various game aspects), and Cluster 2 (reviews on gameplay and technical aspects). The results indicate that the Spectral Clustering method with PCA based on Gaussian Similarity is effective in clustering video game reviews, thereby assisting developers in improving game quality based on identified patterns from the review clusters.