Short-term traffic flow prediction based on IDBO-SVM model

ObjectiveThis study investigated to accurately predict short-term traffic flow, tap the potential value of traffic flow data, and provide a scientific decision-making basis for traffic management departments and travelers. A short-term traffic flow prediction model, IDBO-SVM, was proposed based on the improved dung beetle optimization (IDBO) algorithm and support vector machine (SVM).MethodFirst, the uneven initial population distribution with traditional DBO algorithm was solved by introducing Bernoulli chaotic mapping, avoiding the algorithm falling into local optimum. Second, the adaptive weight factor was added to improve the position update formula of stealing dung beetles; and the global search and local development capabilities of algorithm were balanced. Third, the differential evolution (DE) strategy was integrated to enhance the later convergence ability of algorithm and improve the model accuracy. Finally, eight benchmark functions were selected to simulate and verify IDBO algorithm. Based on the real traffic flow data of M6 motorway in UK, the prediction performance of IDBO-SVM model was compared with PSO-SVM, DE-SVM and DBO-SVM models to verify the effectiveness of the proposed model.ResultThe simulation results indicate that IDBO algorithm has excellent optimization performance in single-peak, multi-peak and fixed-dimensional multi-peak test functions. The optimization speed and accuracy are significantly improved compared with the traditional algorithms. The prediction results show that the MAE of IDBO-SVM model is 23.14 with improvement of 0.11-3.53, the RMSE is 30.79 with improvement of 0.21-4.57, and the MAPE is 3.77 with improvement of 0.005-0.87.ConclusionIDBO-SVM model optimizes SVM parameters through IDBO algorithm in multiple dimensions, effectively overcomes the performance defects of traditional model, and accurately completes the prediction on short-term traffic flow. It has certain application prospects.