Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

User Behavior Analysis and Clustering in a MMO Mobile Game: Insights and Recommendations

Yang Qiu, Yuxin Gong, Guanliang Liu

TL;DR

The paper addresses understanding diverse player behavior and retention in a popular MMO mobile game by integrating temporal/dynamic data mining, graph embeddings, and static attribute clustering. It introduces TimeSeriesKMeans for multivariate time-series clustering, and graph embedding methods DeepWalk and LINE to capture social structure, complemented by hybrid visualizations for interpretation. Key findings reveal five distinct player segments with varying engagement, skill, and social patterns, with Cluster 1 showing strong growth trajectories and server outages exerting negative effects on retention and trust. The work yields data-driven recommendations for personalized experiences, targeted marketing, and robust incident response, offering practical guidance for developers and marketers in online gaming environments.

Abstract

This study presents a comprehensive analysis of user behavior and clustering in a popular mobile battle royale game, employing temporal and static data mining techniques to uncover distinct player segments. Our methodology encompasses time series K-means clustering, graph-based algorithms (DeepWalk and LINE), and static attribute clustering, visualized through innovative hybrid charts. Key findings reveal significant variations in player engagement, skill levels, and social interactions across five primary user segments, ranging from highly active and skilled players to inactive or new users. We also analyze the impact of external factors on user retention and the network structure within clusters, uncovering correlations between cluster cohesion and player activity levels. This research provides valuable insights for game developers and marketers, offering data-driven recommendations for personalized game experiences, targeted marketing strategies, and improved player retention in online gaming environments.

User Behavior Analysis and Clustering in a MMO Mobile Game: Insights and Recommendations

TL;DR

The paper addresses understanding diverse player behavior and retention in a popular MMO mobile game by integrating temporal/dynamic data mining, graph embeddings, and static attribute clustering. It introduces TimeSeriesKMeans for multivariate time-series clustering, and graph embedding methods DeepWalk and LINE to capture social structure, complemented by hybrid visualizations for interpretation. Key findings reveal five distinct player segments with varying engagement, skill, and social patterns, with Cluster 1 showing strong growth trajectories and server outages exerting negative effects on retention and trust. The work yields data-driven recommendations for personalized experiences, targeted marketing, and robust incident response, offering practical guidance for developers and marketers in online gaming environments.

Abstract

This study presents a comprehensive analysis of user behavior and clustering in a popular mobile battle royale game, employing temporal and static data mining techniques to uncover distinct player segments. Our methodology encompasses time series K-means clustering, graph-based algorithms (DeepWalk and LINE), and static attribute clustering, visualized through innovative hybrid charts. Key findings reveal significant variations in player engagement, skill levels, and social interactions across five primary user segments, ranging from highly active and skilled players to inactive or new users. We also analyze the impact of external factors on user retention and the network structure within clusters, uncovering correlations between cluster cohesion and player activity levels. This research provides valuable insights for game developers and marketers, offering data-driven recommendations for personalized game experiences, targeted marketing strategies, and improved player retention in online gaming environments.
Paper Structure (23 sections, 2 equations, 14 figures, 3 tables)

This paper contains 23 sections, 2 equations, 14 figures, 3 tables.

Table of Contents

  1. Data Overview & Overall Analysis
  2. Statistical Analysis
  3. Key Opinion Leader Mining
  4. Data Clustering (Temporal/Dynamic Data Mining)
  5. Feature Engineering
  6. Clustering with TimeSeriesKMeans
  7. Visualization of Clustering with PCA and t-SNE
  8. PCA - Clustered Data Visualization
  9. t-SNE - Clustered Data Visualization
  10. Analysis of Cluster 1's User Profile and Impact of Upcoming Game Content
  11. Impact of Server Issues
  12. User Confidence and Trust
  13. Recommendations for Operational and Public Relations Teams
  14. Static Data Mining
  15. Data Clustering on Graph Structure
  16. ...and 8 more sections

Figures (14)

  • Figure 1: ID(Masked ) of Kols at different time points.
  • Figure 2: PCA visualization of clustered data.
  • Figure 3: t-SNE visualization of clustered data.
  • Figure 4: Trend Changes of Three Clusters.
  • Figure 5: PUBG Official Statement on Login Issues.
  • ...and 9 more figures