Vehicle Travel Time Prediction Based on SSA-LSTM

Highway travel time, defined as the time difference between a vehicle entering and exiting the highway, is primarily influenced by factors such as the highway's characteristics, current road conditions, and the driver's behavior. Travel time is a crucial parameter for assessing the operational state of a highway and serves as an important reference for both drivers when selecting routes and for traffic managers in implementing control measures. Therefore, predicting highway travel time is of significant importance. Long Short-Term Memory (LSTM) is a type of recurrent neural network, an improved version of the traditional recurrent neural network, known for its excellent performance in predicting sequential data, such as time-series data on highway travel times. In operations research, search optimization algorithms have been a key area of focus in conjunction with machine learning. The Sparrow Search Algorithm (SSA), a relatively new search optimization algorithm, boasts high convergence speed and accuracy, along with strong global search capabilities, reducing the risk of getting trapped in local optima. This study utilizes vehicle entry and exit data collected by the ETC system at the Jin-Gang Expressway toll stations to analyze the highway's basic information. Building upon the traditional LSTM model, we integrate the SSA to optimize the LSTM model parameters. By using SSA to search for the optimal parameters, we form a new SSA-LSTM model. Finally, we apply this model to predict travel time data, showing a 120% improvement in prediction accuracy compared to the original LSTM model.