An Improved Multi-Imputation Technique Based on Chained Equations and Decision Trees: Application to Wind Energy Conversion Systems

Missing data (MD) is a prevalent issue that researchers and data scientists frequently encounter. It can significantly impact the quality of analyzed data, affecting the relevance of the interpreted results and the inferred conclusions. In response to this challenge, a novel multi-imputation technique that combines Multivariate Imputation by Chained Equation (MICE) with Decision Tree (DT), namely (MICE-DT), is proposed. This developed method was evaluated against several established imputation techniques, including K-Nearest Neighbors (KNN), K-Means clustering, Decision Tree (DT), and MICE, under the assumption of Missing at Random (MAR). The performance of the MICE-DT algorithm, along with the comparative analysis of the studied techniques, was demonstrated on a Wind Energy Conversion System (WEC), yielding satisfactory results.