In the past decade, researchers have attempted to develop computer-assisted learning and testing systems to help students improve their learning performance. Conventional testing systems simply provide students with a score, and do not offer sufficient information in order to improve their learning performance. It would be of more benefit to students if the test results could be critically analysed and hence learning suggestions could be offered accordingly. This study proposes an algorithm for diagnosing students' learning problems and provides personalised learning suggestions for Science and Mathematics courses. An intelligent tutoring, evaluation and diagnosis system has been implemented based on the novel approach. Experimental results on a Mathematics course have demonstrated the feasibility of this approach in enhancing students' learning performance, making it highly promising for further study. (Contains 5 tables and 5 figures.)