In the core of computational thinking (CT) practices is problem-solving. However, there is little research on connecting CT with problem solving processes. We developed a computer-based assessment on CT that can log students' problem-solving processes and administered it to a group of fifth graders as a pre- and post- measure of a robotics curriculum. In light of our CT framework, we identified certain measures from collected problem-solving process data for each question analyzed and applied basic statistic analysis, clustering algorithms, and visualization techniques to find patterns. We found that students took different patterns in solving a problem even if their total scores were similar. These patterns prove promising in obtaining a better understanding of a student's CT.