Nested Random Forest: A Personalized Imputation Method for Missing Data

In the field of data science, the imputation of missing data is a common and important problem. Accompanied by the development of artificial intelligence, the data analysis methods of machine learning have received unprecedented attention. As one kind of the important algorithms of machine learning, Random Forest has been sought after by researchers for its excellent practical effect and reasonable interpretation. In this paper, a Nested Random Forest imputation method is proposed to deal with the missing data problem. To ensure the authenticity of the filling data, this method not only considers the overall distribution of the data but also pays attention to the personalized characteristics of the data. From the experimental results, the Nested Random Forest algorithm achieves better effects than the seven mainstream baseline methods in three real-world datasets of different application fields.