Automatic network configuration method, device and equipment based on deep reinforcement learning

The invention discloses an automatic network configuration method, device and equipment based on deep reinforcement learning. The method comprises the steps of firstly collecting configuration requirements of a network manager to construct a configuration requirement graph; a network configuration predictor based on a graph attention network (GAT) model is constructed, so that probability distribution of potential parameter values of all unknown parameters is output; when the GAT model is trained, a targeted training strategy is developed to improve the prediction capability of the model. Then, an optimizer based on deep reinforcement learning is designed, and an optimal solution is searched by using a depth deterministic policy gradient (DDPG) algorithm so as to achieve the goal of meeting network specifications to the maximum extent. According to the method provided by the invention, the specification consistency (the degree that the network configuration meets the network specification) of the network conf