An Adaptive Data Preprocessing Framework for Improved Learning: A Case Study of Tangier Container Terminal

Abstract

Container terminals are critical nodes within the maritime transportation system that have a vital function in global merchandise trade, handling a significant volume of cargo through the use of various equipment and personnel. Thus, the efficiency of container terminal operations relies heavily on the ability to collect, analyze, and utilize operational data. However, such data can be corrupted by noise, missing points, outliers, and incomplete or inconsistent information, making subsequent analysis or modeling challenging. This study proposes an adaptive data preprocessing framework tailored to the context of container terminal operations, using data from tangier container terminal as a case study, the leading container port in the Mediterranean and Africa, and also ranked 4^th in the CPPI 2022. This framework includes techniques for data integration, cleaning, transformation, and encoding to acquire high-quality data. In addition, the RFE feature selection method is employed to identify the most discriminative feature subset. Finally, the proposed approach, assessed using an extra tree regressor model, demonstrates strong prediction capabilities with an R-squared score of 95.4% based on the selected features for predicting the duration of vessels at port, highlighting that its integration into the terminal operating system can improve management efficiency.