Imputation of missing values for semi-supervised data using the proximity in random forests

摘要

This paper presents a procedure that imputes missing values by using random forests on semi-supervised data. Applying our method to Hewlett-Packard Lab.'s spam data and Edgar Anderson's iris data, we found that the rate of correct classification is higher than that of other methods: a simple expansion of Liaw's 'rflmpute' for (un)supervised data and the fc-nearest neighbour method (kNN). Our method allows missing predictor variables as well as missing response variable. An imputation that uses random forests for semi-supervised cases in the training dataset has never been implemented until now.