An open-ended serious game can engage students' scientific problem-solving processes. However, understanding how students learn higher-order thinking skills through solving a problem in an open-ended game system is a challenge. The complex game systems may make learning more difficult for students with different characteristics such as an at-risk label. Recent research stresses the importance of using gameplay data to better understand diverse individuals' learning behaviors and performances in the context of serious games. We analyzed gameplay data of a serious game called "Alien Rescue" to identify navigation behavior patterns between at-risk and non-at-risk middle school students. Particularly, we incorporated the combination of lag sequential analysis and sequential pattern mining in statistical analyses. The results revealed that the non-at-risk and at-risk students used problem-solving strategies differently when they navigated the environment. The findings using this integrated method confirmed that additional support was needed for at-risk students in order for them to develop contextual and procedural knowledge for problem-solving in the game environment. The findings provide methodological guidelines for researchers considering a sequential analysis as well as offer practical guidelines for game designers to consider when designing serious games with a complex problem so as to help at-risk students.