|
| ML category | Method | Highlights |
|
| Clustering | Best fit missing value imputation | Used for the IoT datasets. The paper provides a comparison of BFMVI with other existing algorithms and shows that BFMVI outperforms them in terms of accuracy and efficiency [7] |
| Cluster-directed framework for neighbor-based imputation | A brand-new cluster-directed framework is suggested by the authors. CFNI: cluster-directed framework for neighbor-based imputation, which uses data clustering alone to lead the identification of closest neighbors in order to get a more precise imputed value [8] |
| C-means | Used to impute the value in missing places by similar entries in the complete datasets [9]. Used in distributed datasets [10] |
| K-means | Patil et al. used it to impute missing value in their work [11] |
|
| Deep learning | Deep neural network | Able to fit the data closely, and can accurately predict new data points [12] |
| Long-short-term memory | Demonstrates good performance for time series missing values [13] |
|
| Ensemble | AdaBoost | In [14], authors showed that the method is good enough to resilient missing data to identify hemodynamic instability in ICU patients early on |
| eXtreme gradient boosting | Employs feature selection and superior accuracy [15, 16] |
| Random forest | In [17], authors used random forest to estimate categories for similarity measuring to impute missing data |
|
| Neural network | Multilayer perceptron (MLP) | In [18], authors showed the good results in categorical variables using MLP |
|
| Instance based | k-nearest neighbors (kNN) | Pan et al. [19] considered the feature relevance which was measured by their modified KNN |
| Support vector machine (SVM) | It was used to impute missing data for activity-based transportation model [20] |
|
