A Random Forest Model Based on Parameter Optimization Using Cuckoo Search Algorithm for Ship Traffic Flow Forecasting

Abstract

The forecast of ship traffic flow can effectively improve the ability of channel management. One of the difficulties of forecasting lies in the non-linear characteristics of traffic flow distribution over time. This paper proposes a traffic flow prediction model by combining random forest model (RF) and cuckoo search algorithm (CS). CS selects the optimal parameters for the RF model. Then use the RF model optimized by CS to predict the time series of traffic flow. The proposed model has been examined and the convergence of the cuckoo optimization algorithm has been analyzed by using the statistical data of ship traffic flow in Qingdao Port. The results show that, compared with the BP neural network and the random forest model, the CS-RF model is more accurate in predicting ship traffic flow. This can help the channel management office to estimate the ship flow of a port effectively.