Bus arrival time prediction is an important part of urban bus operation, which
maintains the stability and punctuality of the bus system. Providing accurate
public transport travel service information can attract more public transport
travelers, thereby improving the public transport share. By knowing the arrival
time of the bus in advance, travelers can arrange their travel time more
effectively and reduce their waiting time at bus stops. In addition, accurate
bus arrival time also contributes to the design, development and management of
the bus system, promoting better resource scheduling and lower operating costs.
However, the prediction based on historical data cannot cope with the complex
changes of real-time traffic conditions and meet the requirements of public
transportation information system. Therefore, current research is increasingly
focused on improving the accuracy of real-time prediction, while prediction
models are exploring and adjusting to adapt to complex traffic conditions and
real-time changes. In this paper, an innovative prediction method is proposed,
which integrate Complete Ensemble Empirical Mode Decomposition with Adaptive
Noise (CEEMDAN), Temporal Convolutional Network (TCN), and Long Short-Term
Memory (LSTM). This hybrid methodology can be used as a real-time tool to
provide effective short-term bus arrival time prediction without the need of
additional variable input. Finally, a case study was carried out based on the
actual data of Xuzhou Route 1 bus, so the proposed method can be empirically
evaluated and compared with other methods. As a result, the proposed model
offers higher prediction accuracy in bus arrival time prediction, demonstrating
superior performance compared to traditional prediction methods.