In today’s life internet is an important part. We spend most of our time on internet. One of the important features of internet is communication. Email is a mode of communication which is used for the personal and business purpose. Spam emails are the emails recipient does not wish to take delivery of; it is also called unwanted bulk email. Emails are used each day by number of user to converse around the world. At present large volumes of spam emails are reasoning serious trouble for Internet user and Internet service. Such as it degrade user investigate knowledge, it assists transmission of virus in network, it increases load on network traffic. It also misuses user time, and energy for legal emails among the spam. For evade spam there are so many conventional anti-spam technique includes Bayesian based sort, rule based system, IP blacklist, Heuristic based filter, White list and DNS black holes. These methods are based on satisfied of the post or links of the mail. In this paper we proposed an efficient spam filtering technique based on neural network. The technique used is RBF a neural network technique in which neuron are trained. The results obtained by using this technique are compared with SVM. The parameter meter for comparison is precision and accuracy. On the basis of these two parameters we compared the proposed technique with SVM.