To improve the rationality and accuracy of port throughput prediction,an RF-BP model based on multi-variable inputs is established using the improved BP neural network with the random forest(RF)algorithm for fea-ture extraction.The model is used to predict the total throughput of ten major coastal ports in China.Firstly,the RF algorithm is used to extract features from multi-variable input data,which improves the robustness and accuracy of the model by selecting important features.Then,the BP neural network is applied for regression prediction,fully utilizing its fitting and generalization capabilities.Finally,the prediction results are compared with those of single-variable BP neural network model,multi-variable BP neural network model,and ARIMA algorithm.The results show that the multi-variable RF-BP neural network predicts throughput data with smaller errors and more accurate results.The model improves the accuracy of port throughput data prediction,which is significant for enhancing the port·s ability to respond to market demand,supporting port strategic decision-making,optimizing cargo transportation processes and promoting regional economic development.
关键词
吞吐量预测/沿海主要港口/随机森林算法/神经网络/多变量/特征提取
Key words
Throughput prediction/Ten major coastal ports/Random forest algorithm/Neural network/Multi-varia-ble/Extract features
维普
万方数据