In view of the successful application of deep learning, mainly in the field of image recognition, deep learning applications are now being explored in the fields of communication and computer networks. In these fields, systems have been developed by use of proper theoretical calculations and procedures. However, due to the large amount of data to be processed, proper processing takes time and deviations from the theory sometimes occur due to the inclusion of uncertain disturbances. Therefore, deep learning or nonlinear approximation by neural networks may be useful in some cases. We have studied a user datagram protocol (UDP) based rate-control communication system called the simultaneous multipath communication system (SMPC), which measures throughput by a group of packets at the destination node and feeds it back to the source node continuously. By comparing the throughput with the recorded transmission rate, the source node detects congestion on the transmission route and adjusts the packet transmission interval. However, the throughput fluctuates as packets pass through the route, and if it is fed back directly, the transmission rate fluctuates greatly, causing the fluctuation of the throughput to become even larger. In addition, the average throughput becomes even lower. In this study, we tried to stabilize the transmission rate by incorporating prediction and learning performed by a neural network. The prediction is performed using the throughput measured by the destination node, and the result is learned so as to generate a stabilizer. A simple moving average method and a stabilizer using three types of neural networks, namely multilayer perceptrons, recurrent neural networks, and long short-term memory, were built into the transmission controller of the SMPC. The results showed that not only fluctuation reduced but also the average throughput improved. Together, the results demonstrated that deep learning can be used to predict and output stable values from data with complicated time fluctuations that are difficultly analyzed.