MSTIM: A MindSpore-Based Model for Traffic Flow Prediction

TL;DR

MSTIM introduces a MindSpore-based framework that integrates CNN, LSTM, and self-attention to model multi-scale temporal features for traffic-flow forecasting. By applying a multi-scale CNN for local spatial patterns, LSTM for long-term dependencies, and attention to emphasize informative time steps, MSTIM achieves superior accuracy and stability on the MITV dataset, outperforming LSTM-Attention, CNN-Attention, and LSTM-CNN baselines with lower MAE, MSE, and RMSE. The approach demonstrates effective capture of complex spatio-temporal dynamics, yet remains limited to a univariate, single-road setting and leaves open efficient deployment and multi-source data fusion for future work. Overall, MSTIM offers a practical, robust solution for traffic-flow forecasting with potential impact on ITS planning and dynamic traffic management.

Abstract

Aiming at the problems of low accuracy and large error fluctuation of traditional traffic flow predictionmodels when dealing with multi-scale temporal features and dynamic change patterns. this paperproposes a multi-scale time series information modelling model MSTIM based on the Mindspore framework, which integrates long and short-term memory networks (LSTMs), convolutional neural networks (CNN), and the attention mechanism to improve the modelling accuracy and stability. The Metropolitan Interstate Traffic Volume (MITV) dataset was used for the experiments and compared and analysed with typical LSTM-attention models, CNN-attention models and LSTM-CNN models. The experimental results show that the MSTIM model achieves better results in the metrics of Mean Absolute Error (MAE), Mean Square Error (MSE), and Root Mean Square Error (RMSE), which significantly improves the accuracy and stability of the traffic volume prediction.

Figures (3)

• Figure 1: Structural diagram of the MSTIM traffic flow prediction model, which integrates convolutional neural network (CNN), long short-term memory network (LSTM) and attention mechanism to realize accurate prediction of future traffic flow.
• Figure 2: Structure of an LSTM cell. The diagram illustrates the internal flow of information, including the forget gate ($f_t$), input gate ($i_t$), candidate memory cell ($\tilde{C}_t$), output gate ($o_t$), and the updates of long-term memory ($C_t$) and hidden state ($h_t$).
• Figure 3: A schematic diagram of a convolutional neural network (CNN), illustrating the typical flow from input through convolutional and pooling layers to fully connected layers and final output.