This study uses univariate ARIMA, neural network, and GM (1,1) models to predict the container throughput in Kaohsiung port and uses MAPE, RMSPE and THEIL evaluation index to evaluate their efficacy.

• 本研究以单变量ARIMA模式、类神经网路及灰色GM(1,1)模型三种预测模型应用于高雄港货柜吞吐量之预测，且透过MAPE、RMSPE与THEIL三种评鑑指标加以评估其效能。