The fast growth of E-learning around the world is inspiring many educational and business institutions to adopt the adaptive personalized method of teaching and learning. However the success of E-learning does not depend only on the academic setup but also on proper realization of its vital requisites. This paper presents an idea for building recommendation system for the e-Learning system using Association Rule Mining Techniques for the best selection of e-learning resources or learning materials . This study analyzes students' log of a Learning Management System (LMS) Moodle, and data gathered from a survey dataset addressing students academic, interaction and personal information. Data mining and statistical tools have been used to find relationships between students' LMS access behavior, study habits and overall performances. Our investigation shows that majority of the students feel LMS as 'very helpful' tool as teaching-learning process and students having low LMS access rate obtained poor grade. But due to unattractive LMS interface and slow Internet speed, more than half of the students still prefer the instructor to provide the printed lecture notes. The use of association mining and the use of distance metrics such as the Jaccard measure to identify LMS that belong to the same genre will be used for constructing the recommendation System for the best selection of learning resources. The efficacy of teaching learning will be increased due to this association based recommendation system. © EuroJournals Publishing, Inc. 2011.