A novel method for estimating missing values in ship principal data

Abstract

Missing values in the fleet data set acquired in the marine sector reduce the data available for analysis, which can decrease the statistical power of the model and negatively affects the energy-efficient operation and decision-making. This article presents a method to estimate ship principal data. A model-based computation method using regression analysis was used to handle missing values, and a case study was conducted on principal data from 6,278 container ships in the IHS Sea-Web database. To implement a model for predicting missing values, the entire data set was randomly divided into 80% to 20%, which were used as a training data set and test data set. The prediction performance of models was compared with several regression equations proposed in prior studies, which shows that there is a significant improvement with our method. The goodness of fit of the current method has increased by up to 15.6% over the previous methods. It also showed good applicability for ships with restrictions on certain dimensions, such as the standards for Suez and Panama Canal. The findings presented here may be helpful from the estimation for key parameters of the ship to the computation of missing values in the marine sector.