This article is concerned with the prediction of time delays and a control method for networked control systems (NCSs). We propose to predict the network-induced delays online with least squares support vector machines (LSSVM) optimized by particle swarm optimization (PSO). Then self-tuning proportional–integral–derivative (PID) control with variable delays is adopted in the NCSs. The recursive least squares (RLS) method is used to identify the parameters of the controlled objects, which can adapt to changes in dynamic models. Simulations verified the excellent performance of network delay prediction and networked control, and stable closed-loop control can be realized.