A process for analyzing log files collected from open-ended learning environments is developed and tested on a virtual lab problem involving reaction stoichiometry. The process utilizes a set of visualization tools that, by grouping student actions in a hierarchical manner, helps experts make sense of the linear list of student actions recorded in a raw log file. Such analysis of an initial set of log files is then used to develop a rule-based system that can automatically classify the problem-solving strategies being engaged in by the students. The strategies assigned by the resulting rule-based system compare well with strategy codes assigned by experts.