Machine Learning-Based Predictive Analysis for Assessing Operational Efficiency in Indian Port Management Systems

Abstract-- Port operations in India are becoming increasingly complicated as vessel arrivals, cargoes, and weather conditions change, necessitating the use of data-driven decision-making to enhance efficiency. Existing systems based on rule-based logic and disconnected deep learning models suffer from irrelevant, unchangeable thresholds, poor explainability, and isolated data
connections, resulting in inaccurate stay time and berth
assignment. The proposed machine learning-based predictive
analysis platform uses AIS, manifest data, meteorological inputs, and operational logs to estimate vessel berth usage, delay class, and berth productivity. Using 2019-2024 Indian port datasets for training, it achieves an MAE of 4.1 hours, an 18% improvement over the old system, increases the delay classification F1-score from 0.68 to 0.79, and provides a 9.4% gain in berth productivity. These advantages include more accurate prediction, increased decision
making transparency, and quantifiable operational efficiency gains in various maritime conditions.
Keywords: Predictive Analytics, Port Management, Operational
Efficiency, Dwell Time Prediction, Berth Productivity, Maritime
Logistics, Decision Support System.