information for data mining.
patterns of behavior, structure or content in
order to improve the overall gameplay, hence
keeping players longer and increasing the
revenue of the game service.
describe it from web usage mining perspective,
with the main emphasis on data gathering.
A classification of game types from a data
mining perspective is also suggested.
common game log format suitable for game
usage mining.
the traditional Internet, are excellent sources
of information for data mining. We can mine
usage patterns of both human and virtual
players (avatars), or discover patterns in the
games content and structure.
massive multiplayer games (game usage min-
ing) to the similar process known from web us-
age mining.
ing in computer games is to discover patterns,
e.g. rules or statistics, that can possibly be
used to improve the game. Then paying players
become more satisfied and stay longer, which
again increases the revenue of the game ser-
vice. This is of particular importance for wire-
less games, since keeping a player is equiva-
lent of making more money. Wireless Inter-
net revenue models are either proportional to
money per time unit spent by the player (e.g.
WAP over GSM), or increasingly more com-
mon, money per byte served to the player (e.g.
UMTS, I-Mode and WAP over GPRS).
tures in games. From a web usage mining per-
spective they can be seen as relatively analog to
web pages and links (i.e. URLs), respectively.
Player behavior can also be detected, including
actions and speech act utterances. This makes
it possible generate game logs that resemble
web logs in structure, and hence can gain from
applying web usage mining methods with only
minor adaptions.