Traffic prediction plays a crucial role in improving road network utilization and reduces travel delays in urban areas. This complex dynamic problem has attracted many scholars and practitioners to join in solving the task. This study aims to introduce a new approach to predict short-term traffic conditions by using a variant of deep learning techniques. Specifically, the proposed method applies a combination of a diffusion model with a residual convolutional recurrent neural network to address the spatiotemporal characteristics of the road network. To improve the prediction performance, a Gaussian optimization was employed to guide the parameter tuning during the training process. The method was applied to solve a traffic prediction problem in Jakarta city, the capital of Indonesia. The computational results suggest that the proposed method is highly competitive and provides a significant improvement compared to the existing methods in the literature
