A novel data condition and performance hybrid imputation method for energy efficient operations of marine systems

A novel framework for imputing large gaps of missing values from time series sensor data of marine machinery systems

“…In addition, for validation purposes, the performance of the proposed framework is compared with one of the most widely implemented imputation techniques, the Multivariate Imputation by Chained Equations (MICE), as presented in Cheliotis et al (2019).…”

“…In almost all the studies, with the exception of one that considered structured data of patients, the imputation techniques were implemented in time series (e.g., Pratama et al 2016;Priya Stella Mary and Arockiam 2017;Bashir and Wei 2018;Azimi et al 2019;Hadeed et al 2020) to tackle the challenges derived due to the application of IoT. In the maritime industry, the analysed time series collected from sensors coupled to marine machinery systems contained between 4.4% and 26% missing values (Cheliotis et al 2019). Consequently, if incomplete time series are not addressed appropriately, the resulting data analysis may be unreliable and inaccurate, which yields bias in succeeding steps, and thus leads to poor models being implemented to assist decision-making processes (Fekade et al 2018).…”

“…Furthermore, there is no evidence of any study that presents a novel univariate imputation method to deal with the missing data from time series sensor data of marine machinery systems. In this regard, only one study has been identified in this sector, in which the implementation of k-NN model in conjunction with the MICE method was considered to impute missing values (Cheliotis et al 2019). However, both imputation techniques are multivariate methods, and thus predictors are required to complete the instances of a parameter.…”

Optimized SVM-Driven Multi-Class Approach by Improved ABC to Estimating Ship Systems State

“…Similar to (18), the equivalent Lagrange dual function of (27) can be obtained via the Lagrange multiplier method. (27) where α i ≥0, μ i ≥0 are Lagrange multipliers.…”

“…Zhang et al [17] used a single-parameter threshold method and radar chart analysis method to evaluate the health status of ships' power equipment, and converted the evaluation results into health score values. Cheliotis et al [18] introduced a data condition and performance hybrid imputation method for energy efficient operations of marine systems, which combined data mining with first-principle knowledge and was implemented in a dataset acquired from the main engine system of an oceangoing vessel. Zhang et al [19] established a real-time state evaluation system for the power quality of an electric propulsion ship using the fuzzy analytical hierarchy process method, which strengthened the supervision for the ship electric power system.…”