Feature-driven Data Science in Game Development
Framework and methodology to utilize Data Science to augment better and more profitable game development.
I designed and implemented a system that uses feature-driven approach to restructure games and their mechanics into recognizable general actions linked to player behavior, which we formulate as standard universal features. Framework collects data and analyzes it to recognize behavior patterns and player groups in the context of game features.
System produces statistics and data analytics of past behavior, but its primary purpose is to provide predictive analytics for new games under development. When any game and its mechanics are described with standard universal features used in previous games, we have an opportunity to use past player behavior data from other games to predict how would players react to a new game (which is under development) with certain set of features using Bayesian Inference and other Data Science methods.
Additionally, user can simulate game monetisation in similar manner for IAP, DLC and other monetisation categories.
Technology used in data collection and pipelines: Node.js, ArangoDB, Apache Spark.
Machine learning implemented in: R and Python
Dashboards are implemented in: Node.js, Express, Angular, D3.js, dc.js, crossfilter, Three.js
You can read more about this in my LinkedIn article: