One of the research areas in intelligent transportation is traffic prediction. In order to more thoroughly examine the temporal and spatial characteristics of traffic flow sequences and increase forecast accuracy, this study suggests a short-term traffic flow prediction model based on discrete wavelet transformation (DWT) and graph convolutional networks (GCNs). First, the DWT method breaks down the original traffic sequences into precise and approximative components in order to lessen the non-stationarity of the traffic flow data. In order to extract the spatial properties of road networks, the adjacency matrix of the GCN model is secondly improved by the addition of the distance factor term. Finally, the prediction results from each group of components deconstructed by DWT are combined to provide the final prediction value, which is fed to the GCN model independently for each group. The proposed model's mean absolute error (MAE) and mean absolute percentage error (MAPE), when compared to the ARIMA, WNN, and GCN models, have been reduced by $57 \%$ and $59\%$, respectively. This indicates that the proposed model is a successful method for forecasting short-term traffic flow. The model is tested on the Caltrans PeMS dataset.