User Disengagement Prediction Model for Online Video Games
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Users of computer gaming systems often disengage from videogames, leading to a need for methods and systems that can effectively retain user engagement.
Innovation Solution
A computing system and method that extract raw user data from various sources, segment and process it to derive attributes, impute missing data, and use natural language processing to classify unstructured data. This processed data is then used to train a prediction model to anticipate user disengagement, allowing for targeted interventions to reengage users.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If game developers design games to provide intrigue, challenges, and achievement opportunities, then user engagement is improved, but users still disengage over time
Solution Approach 1:
The system performs preliminary actions by continuously collecting user data, deriving attributes, and training prediction models in advance to identify users who are likely to disengage. This allows the system to proactively intervene before actual disengagement occurs, sending notifications and offers to retain users early in the disengagement process rather than waiting for them to leave completely.
Solution Approach 2:
The system implements feedback loops by continuously monitoring user engagement metrics, updating prediction models with new data, and adjusting retention strategies based on model predictions. The system sends notifications to users predicted to disengage and tracks whether these interventions successfully retain users, using this feedback to continuously improve the prediction model and retention effectiveness.
2Measurement precision
If the system collects and processes extensive user data to improve prediction accuracy, then prediction precision is improved, but data processing complexity increases
Solution Approach 1:
The system segments user data into distinct categories and derives specific attributes from different data sources. User data is divided into engagement metrics, transactional data, demographic information, and behavioral patterns. Each segment is processed separately through specific derivation rules, making the complex data processing task more manageable and systematic while maintaining high prediction accuracy.
Solution Approach 2:
The system introduces an intermediary layer of data derivation and attribute extraction between raw user data and the prediction model. Instead of feeding raw data directly into the model, the system processes data through intermediate steps that derive meaningful attributes (e.g., engagement score, churn risk indicators) which then serve as inputs to the prediction model, simplifying the overall processing complexity.
3Reliability
If the system implements comprehensive data derivation and processing to identify user disengagement patterns, then user retention capability is improved, but computational resources increase
Solution Approach 1:
The system applies partial action by focusing computational resources on processing data for users who are most likely to disengage, rather than uniformly processing all user data at the same level. The prediction model identifies high-risk users, and the system intensively processes and monitors this subset of users with more sophisticated analysis, while using lighter processing for low-risk users, thereby reducing overall computational resource usage while maintaining effective retention capability.
Data Source
AI summary
Systems and methods extract raw user data of users of an online video game application (OVGA) and append raw user data to a dataset. Data object(s) of the dataset are segmented into, and attribute(s) are derived from field(s) of the dataset, and the attribute(s) include a time between initiation of user engagement with the OVGA and user disengagement with the OVGA by the users. Missing data is imputed with substitute values, and entity extraction is performed using natural language processing to extract specific data that is then classified and stored to a relational database. The relational database is versioned to generate versioned data, and a target variable value is inserted into an iterative training and testing loop to train a prediction model using the versioned data, where the training is based on adjusting weights and iterative calculations to predict the a date of a churn-based event.