This study proposes a sequence alignment-based behavior analysis framework (SABAF) developed for predicting inactive game players that either leave the game permanently or stop playing the game for a long period of time. Sequence similarity scores and derived statistics form profile databases of inactive players and active players from the past. SABAF uses global and local sequence alignment algorithms and a unique scoring scheme to measure similarity between activity sequences. SABAF is tested on the game player activity data of Ever Quest II, a popular massively multiplayer online role-playing game developed by Sony Online Entertainment. SABAF consists of the following key components: 1) sequence alignment-based player profile databases, 2) feature selection schemes and prediction model building, and 3) decision support model for determining inactive players.