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

Yang Qiu,Yuxin Gong,Guanliang Liu

2024-09-26

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.

Social and Information Networks

What problem does this paper attempt to address?

The paper aims to address the following issues:

1. **User Behavior Analysis**: By analyzing the behavior patterns of players in large-scale Massively Multiplayer Online (MMO) mobile games, the characteristics and preferences of different player groups are revealed. The study employs time series and static data mining techniques to identify different player segments.

2. **User Clustering**: Utilizing time series K-means clustering, graph algorithms (such as DeepWalk and LINE), and static attribute clustering methods to classify players, in order to better understand the behavioral characteristics of different player groups.

3. **Impact of External Factors**: Analyzing the impact of external factors (such as server failures) on user retention and network structure, exploring how these factors affect player activity and social interactions.

4. **Personalized Gaming Experience and Marketing Strategies**: Based on the above analysis results, providing data-driven recommendations for game developers and marketers to optimize personalized gaming experiences, formulate targeted marketing strategies, and improve player retention.

Specifically, the paper addresses these issues through the following methods and steps:

- **Data Overview and Overall Analysis**: Including statistical analysis and Key Opinion Leader (KOL) mining, using the PageRank algorithm to calculate user influence.
- **Data Clustering (Time/Dynamic Data Mining)**: Selecting appropriate attributes through feature engineering, clustering users using the time series K-means algorithm, and visualizing clustering results through PCA and t-SNE techniques.
- **Static Data Mining**: Clustering based on users' static attributes and using mixed charts (radar charts and violin plots) to show differences between different clusters.
- **Network Attribute Analysis**: Analyzing the network attributes of different user groups (such as connected components, average clustering coefficient, and the number of triangles) to assess the cohesion and vulnerability of each group.

Through these methods, the paper provides in-depth insights into player behavior and social interactions, offering valuable references for game developers and marketers.