Analisis Segmentasi pada Pengelompokan Game di Platform Steam Menggunakan Hybrid K-Means PSO dan FP-Growth

Abstract

The game industry on the Steam platform is characterized by intense competition, making it difficult for developers, especially those from indie community, to understand market dynamics. This research aims to perform a quantitive market segmentation of games on Steam to identify distinct market segments and uncover combination of features (tags) that characterize each segment. The research methodology consists of two main analytical stages. The first stage involves segmentation using a Hybrid K-Means algorithm optimized with Particle Swarm Optimization (PSO) based on two quantitive variables: Review Volume and Positive Ratio. The second stage applies the Frequent Pattern Growth (FP-Growth) algorithm to each resulting segment to analyze the association patterns among tags. The clustering results succesfully identified three statistically and characteristically different market segments: Segment 0 (Mainstream Positive), Segment 1 (Negative & Niche Reception), and Segment 2 (High Review Volume). Further association analysis revealed that each segment possesses a unique “formula” of feature combinations. Segment 0 is dominated by a “hub-and-Spoke” pattern where the Adventure genre serves a foundation, enhanced by strong narrative (Story Rich) and aesthetic (Pixel Graphics) elements. Segment 1 exhibit a fragmented pattern with generin and non-synergistic genre combinations, indicating standard executions that fail to stand out. Meanwhile, Segment 2 shows several distinct clusters of strong, synergistic features, such as the Building + Sandbox + Simulation combination or the Action + Shooter + PvP / Co-op combination, representing formulas for blockbuster games. This research successfully maps the market structure of Steam games and provides actionable, strategic insights regarding the feature combination that correlate with success in each market segment.