Virtual Learning Environments (VLE) have become popular in higher education in recent years due to their ability to provide additional and flexible solutions for students and researchers. However, the limitations of VLEs have led to the development of a new generation of VLE – the Personal Learning Environment (PLE). PLEs avoid these limitations and have new features that allow students to control and develop new applications, such as Web 2.0 applications and social networks. Whilst PLEs have resolved some of the drawbacks of VLEs, it is argued that PLEs also have greater potential to cover a wider range of aspects. This paper presents a proactive context-aware architecture for PLE supporting two major objectives: lifelong access and learner-centric study, covering both traditional formal (institution-based) and informal (private, non-institution-based) academic learning. Bayesian Networks are graphical modeling tools that have been used for modeling uncertain knowledge. Moreover, BN has been used in this research to implement the proposed architecture.