Network-induced delays pose a significant challenge for high-performance disturbance suppression in networked control systems (NCSs), particularly when compounded by data packet losses. To address the challenge, this article proposes a data-prediction-based equivalent-input-disturbance (DPEID) method. First, the impact of short delays is modeled as a delay-induced disturbance and treated as part of a lumped disturbance. Second, a DPEID estimator is developed that utilizes a long short-term memory (LSTM) neural network-based predictor to improve the disturbance-suppression capability in the presence of packet losses. Meanwhile, a radial basis function (RBF) neural network is integrated into the observer to approximate the system uncertainty, thereby improving the estimation accuracy of system states and the disturbance. Furthermore, a dual-structure control strategy is constructed based on a DPEID estimator and a networked predictive controller. This design enables the system to achieve satisfactory control performance despite the communication constraints. Finally, the effectiveness and superiority of the algorithm are verified by experimental results.