Conducting Vessel Data Imputation Method Selection Based on Dataset Characteristics

Abstract Time series datasets collected from marine sensors inevitably undergo missing data problems. This cause unreliable sensor to assist the decision-making process. Many methods are offered impute values. However, selecting best imputation method is not a trivial task, as it usually requires domain expertise and several trial-and-error iterations. Furthermore, when imputations carried out in careless way, generates high error factor that can lead stakeholders wrong assumptions. paper provides systematic approach able extract characteristics of underlying and, based on it, recommends less error-prone method. We evaluate our proposed using nine real-world vessel datasets. In total, we generated 3859 samples consisting 17 inputs 1 target feature. Experimental results show capable obtaining weighted F1-Score 92.6%. Additionally, compared with application selected methods, work gain up 86% average score, worst case being 5%. empirically demonstrate efficient methods.