Data stream mining becomes more and more critical nowadays. These algorithms can be used in various scenarios in the industry and daily use. Dynamic networks, user-generated data, and so many other causes are the main sources of data streams. Hoeffding Trees as a data stream classification approach have shown a good performance in dealing with the data stream related challenges. This method makes the any-time prediction possible. Concept drift is a well-known problem coexist with data streams and there are plenty of approaches available today that tried to handle this problem. In this article, we introduced a new algorithm that uses VFDT and neural network to handle concept drift phenomenon in an acceptable way. This algorithm also benefits from fast startup ability which helps systems to be able to predict faster than other algorithms at the beginning of data stream arrival or concept drift event. We also have shown that our approach will overperform other controversial approaches for the classification task.