Multiple imputation of maritime search and rescue data at multiple missing patterns

Area of Science:

• Data Science
• Maritime Operations Research

Background:

• Maritime search and rescue (MSAR) operations generate critical data that often suffers from missing values.
• Incomplete datasets hinder effective data analysis and decision-making in time-sensitive MSAR scenarios.

Purpose of the Study:

• To assess the efficacy of multiple imputation techniques for handling missing data in MSAR datasets.
• To compare the performance of Data Augmentation (DA) and Expectation-Maximization with Bootstrap (EMB) algorithms under various missing data conditions.

Main Methods:

• Multiple imputation techniques were applied to construct complete datasets based on different missing data patterns.
• Probability density curves and overimputation diagnostics were employed to evaluate the imputation effects.
• Comparative analysis of Data Augmentation (DA) and Expectation-Maximization with Bootstrap (EMB) algorithms was conducted.

Main Results:

• Data Augmentation (DA) demonstrated high operational efficiency but struggled with high data missing rates.
• The EMB algorithm effectively restored dataset distributions across varying missing rates and positions.
• Overimputation diagnostics proved valuable for assessing imputation quality and inter-dataset correlations.

Conclusions:

• The EMB algorithm is a robust method for imputing missing MSAR data, particularly when missingness is high.
• While DA is efficient, its applicability is limited by high data missing rates.
• Overimputation diagnostics offer deeper insights for data mining and improving imputation strategies.