Forecasting of Road Traffic Flow Based on Harris Hawk Optimization and XGBoost

Abstract

With the development of society and economy, people's living standards are improving day by day. The number of private cars is increasing, and the problem of urban traffic congestion is becoming more and more serious. Short-term traffic flow prediction is crucial to assist intelligent transportation system decision-making, solve congestion problems, and improve road capacity. In order to effectively improve the prediction accuracy and improve the generalization performance of the model, this paper combines extreme gradient boosting (XGBoost) and harris hawk optimization (HHO) to propose a multi-step prediction hybrid model. When building a hybrid model, the hyperparameter selection of the XGBoost model is converted into an optimization problem, and the optimization problem is solved through HHO. The solution to the final optimization problem is the optimal parameter combination of the XGBoost model. In order to verify the performance and competitiveness of the model, this study applies the proposed model to traffic flow prediction together with seven other representative models. The results show that the model has high accuracy and stability in practical applications.

Application of Stacking-Based Ensemble Learning Model for Water Quality Prediction

“…They formed an integrated Bayesian model using the results of SWAT-VSA, ANN, and ARMA [11]; Jiayue Gu et al proposed an integrated learning model with KNN, XGB, SVR, and ANN as the base models. They used this model to predict monthly rainfall and finally obtained the prediction results with high accuracy [12]; Longfeng Zhang et al proposed a model combining the Harris Hawk optimization algorithm with the XGBoost model and predicted the short-term traffic flow, and the results showed that the accuracy and stability of the model were more significantly improved [13].…”