Forecasting port container throughput is crucial due to its impact on economic development. Socio-economic factors, which introduce uncertainty, are increasingly integrated into throughput forecasting. The complexity of common multivariate forecasting models significantly affects accuracy, yet few studies compare their performance on the same time series for throughput modeling. This study implements, evaluates, and compares the performance of eight multivariate forecasting models for port throughput within a proposed multiple-input single-output (MISO) system, chosen for their frequent use in container throughput research. It investigates two data preprocessing approaches: Random Forest Variable Importance Method (RF-VIM) and a Multi Lagged Value approach. The comparison uses six error metrics: normalized root mean squared error, mean absolute error, mean absolute percentage error, mean error, and root mean percentage error. Performances are discussed, and recommendations for adopting a suitable model are provided.