Contemporary web browsers do not provide customized recommendations for the users; rather than some suggestions based on cookies or browsing history after content filtering. Usually, most of the users provide some key words to search the contents inside their preferred websites and based on these key words web servers provide the contents. So, it would be an interesting and important idea to provide suggestions or recommendations for the preferred websites based on the user's emotional behaviours and preferences. In this paper, we proposed an efficient recommendation system by means of an emotional web browsing agent which not only can take input from the environment accurately but also provides efficient recommendations of the ranked websites based on the training datasets and inputs. Extensive experimental results show that the proposed emotional web browsing agent can recommend specific websites based on user's emotions very efficiently.