When you are destined for an important appoint-ment, you would obviously opt for the most reliable route instead of the shortest in order to be well prepared". Modern networking is presently undergoing through a quantum leap. To cope up with ambitious demands and user expectations, it is becoming more complex both structurally and functionally. Software Defined Networking (SDN) happens to be an instance of such advancements. It has significantly leveraged the network programmability, abstraction, and automation. Eventually, with acceptance form all major network infrastructure such as 5G and Cloud, SDN is becoming the standard of future networking. Likewise, Machine Learning (ML) has become the trendiest skill-in-demand recently. With its superiority of analyzing data, makes it applicable for almost every possible domain. The attempt to applying the power of ML in networking has not been too long, it allows the network to be more intelligent and capable enough to take optimal decisions to address some of its native problems. This gives rise to Self-Organized Networking (SON). In this article, Routing using Deep Neural Network (DNN) on top of SDN is addressed. We proposed a Self-organized Knowledge Defined Network (SO-KDN) architecture and an intelligent routing algorithm, that reactively finds the most reliable route, i.e., a route having least probability of fluctuation. This reduces network overhead due to re-routing and optimizes traffic congestion. Experimental data show a mean 90% accurate forecast in reliability prediction.