The variations of cable forces directly reflect the internal mechanical states of a cable-tayed bridge. Therefore， the monitoring of stayed cables is important for health evaluation of the cable-stayed bridge. However， most of the existing research on stayed cables focuses on force identification. The prediction of future forces based on the historical data is still difficult to achieve.Therefore， this study proposes a cable force prediction method using the gated recurrent unit （GRU） neural network. The application framework based on the GRU neural network is established by using the processing ability of the GRU neural network to time series data， taking into account the strong serialization characteristics of cable force data. The network construction includes the input layer， the GRU hidden layer and the output layer. The cable stress time history data of a real-world cable-stayed bridge is collected as the training and validation samples for the GRU neural network. Data slices and normalization are applied to the sampling process. The GRU neural network is successfully established to predict future cable force of this bridge. The Network calculation is performed using a gradient descent optimization algorithm. The analysis results show that the proposed method can provide satisfactory predictions for cables of different lengths.