| Title |
Multivariate Imputation by Mahalanobis Distance Optimization (MIMDO) |
|
Authors |
Labita, Geovert John D., Labita, Altea S. and Tubo, Bernadette F. |
|
Publication date |
2025/06 |
|
Journal |
Reliability: Theory & Application (RT&A) |
| Volume |
20 |
| Pages |
632-641 |
| Publisher |
Gnedenko Forum Publications |
| Abstract |
This paper introduces a new method for missing data imputation based on an optimization approach and is now available as an R package called "mimdo". This method deals with imputing the missing values by computing the values that minimize the Mahalanobis distance between an observation and the overall mean. The effectivity of mimdo was demostrated in both classification and regression tasks using popular benchmark datasets. From all experiments, it was found out that using mimdo for imputing the missing values in the dataset, on the average, the classification rate is more than 80% and an R-squared of more than 50%. Furthermore, the consistency of the results were validated through simulation studies. |
| Index terms / Keywords |
missing data, cluster analysis, regression analysis, optimization, mahalanobis |
| DOI |
|
| URL |
|
Mindanao State
University - Iligan Institute of Technology