اعداد نظام توصية لبيئة تعليمية اعتماداً على تحصيل الطلاب في مخرجات عملية التعلم
عبير حسن عبد الرحيم دار موسى
ABEER H A dar mousa
Students in any learning environment differ in their level of knowledge, achieved learning outcomes, learning style, preferences, misunderstand and attempts in solving and addressing problems when their expectations are not met. When a student searches the web as an attempt to solve a problem, he suffers from the large number of resources which are, in most cases, not related to his “needs”, or may be related but complex and advance. The result of his search might make him more confused, scattered, depressed and finally result in wasting his time which – in some cases -may have negative effects on his achievements. From here comes the need for an intelligent learning system that can guide studentsbased on their needs. This research attempts to design and build an educational recommender system for a web-based learning environment in order to generate meaningful recommendations of the most interested and relevant learning materials that suit students’ needs based on their profiles1 . This can be achieved by accessing students’ history, exploring their learning navigation patterns and making use of similar students’ experiences and their success stories. The study proposed a design for a hybrid recommender system architecture which consists of two recommendation approaches: the content and collaborative filtering. The study concentrates on the collaborative recommender engine which will recommend learning materials based on students’ level of knowledge, looking at active students' profiles, and achievements in both learning outcomes and learning outcomes levels making use of similar students’ success stories and reflecting their good experience on active student who are in the same level of knowledge. The design of the collaborative recommender engine includes the “learning” module from which the engine learns past students’ access pattern and the “advising” module from which the engine reflects the experience of similar success stories on active students. The content base recommender engine with its suggested stages is considered as future work, the research used the k-mean cluster algorithm to find out similar students where five distance function are used: Euclidean, Correlation. Jaccard,cosine and Manhattan. The cosine function shows to be the most accurate distance function with the minimum SSE but the highest processing time that doesn’t differ a lot when compared the rest functions. The best number of clusters for the selected dataset was determined using three methods Elbow, Gap-statistic and average Silhouette approach where the best number of cluster shows to be three. The research used the two result rating matrices of similar good and good students with Learnings material in order to calculate learning material weights and rank them based on highest weights which results in a final recommendation list.
علم الحاسوب , Computer Science