PREDICTIVE ANALYSIS OF OPERATIONAL RISK IN CONTAINERS AT THE PORT OF SANTOS

This paper proposes a predictive operational risk model to prioritize 
container inspections at the Port of Santos. Using a synthetic dataset built from 
realistic parameters, we applied preprocessing (One-Hot, scaling), PCA for 
dimensionality reduction, K-Means clustering and supervised classifiers (KNN, 
SVM). The SVM achieved 93.4% accuracy on stratified validation. We discuss 
practical implications, limitations of synthetic data and pathways for integrating 
the model into port operational routines.