Sentence examples for imputed features from inspiring English sources
Table 5 shows how the best 6 classifiers (marked by bold font in Table 7) perform on imputed features when using each MR modality separately.
The lowest error when using complete features is around 0.14 SVM-mlp 0.136 SVM-mlp 0.136e use imputed features the lowhileerror if 0.216 (dLDA).
Tables 11, 12, and 13 from the Appendix list the performance of the best supervised classifiers (marked by bold font in Table 7) when using, respectively, perfusion, diffusion, or spectroscopy data separately, considering imputed features only.
A first conclusion that we can draw from a comparative analysis of the different classifiers is that we obtain the lowest error when training classifiers on data with complete features and not on data with imputed features, no matter the imputation method (our in-house method or the built-in method).
We further observe that the importance of the imputed features in the learned classifiers is significant, implying that their imputation is important for prediction.
Techniques like mean-value imputation can produce models that are biased towards predicting all data with the imputed features as 'negative'negative
With little or nothing known about the ancestral forms of our closest relatives, we can at best impute their features using the modern species: a top-down approach.
Imputing the GO features: We used the above approach to impute GO annotations in both Salmonella and human proteins.
[27] Missing (software flagged features) values were imputed by the K-nearest neighbor method, and genes associated with tumors vs. normal were determined with the PAM algorithm, using the pamr Bioconductor package.
Methods that use the limited available features to impute a large number of missing features are likely to produce very biased models.
Specifically, features with high percentage (>90%) of missing values are omitted from our analysis, whereas the values of the features with less percentage of missing values are imputed with the modes and means, in the case of nominal and numeric features, respectively.
