Researchers have found that students might get lost or feel frustrated while searching for information on the Internet to deal with complex problems without real-time guidance or supports. To address this issue, a web-based collaborative learning system, Collab-Analyzer, is proposed in this paper. It is not only equipped with a collaborative mechanism for helping group members solve problems on the web, but also facilitates teachers and researchers in analyzing students' collaborative information-searching behaviors and performance by providing eighteen quantitative indices. To examine the effectiveness of the proposed learning system, a total of 224 university students and 16 teachers participated in an experiment, in which a web-based problem-solving activity for a social studies course was conducted. The experimental results show that the teachers gave positive evaluations of Collab-Analyzer in terms of promoting students' web-based problem-solving ability and information searching skills. Moreover, the students showed high agreement with both the usefulness and ease of use of the system. The research findings also show that high interactive type students have more significant learning performance than moderate and low interactive students. Finally, implications and suggestions for how to accommodate the needs of students after adopting the proposed Collab-Analyzer system are given.