AAAI Publications, Twenty-Third International FLAIRS Conference

Font Size: 
A Largest Common Subsequence-based Distance Measure for Classifying Player Motion Traces in Virtual Worlds
Nikhil S. Ketkar, G. Michael Youngblood

Last modified: 2010-05-06

Abstract


As first-person based computer games and online virtual worlds become increasingly popular, the problem of developing classification models for player motion traces becomes increasingly important.The key applications of developing such models includes building player profiles for adaptive game play, identifying cheating in theform of bots and detecting, and curbing gold farmer activity. In this work we study the problem of developing classification models for player motion traces in virtual worlds.Specifically, we introduce a distance measure for player motion traces based on computing the largestcommon subsequence between two player motion traces. We evaluated our approach on a large corpus of player data consisting of a total of 45 binary classification problems. The kernel function based on LCS distance measure significantly out performed the Bag of Words kernel on 27 of the problems.

Full Text: PDF