Table 9 presents a summary of the informative attributes and their ANOVA analysis results.
Most of the informative attributes were related to unit-hour analgesic consumption (e.g., PCA analgesic consumption in the 9th hour (pcadose_9hr)) (Table 8).
These findings have suggested that the predictive power of machine learning models in assessing functionally important mutations can be significantly enhanced by selecting informative attributes characteristic of a specific protein family.
These results indicate the significance of the more important and informative attributes identified for the 72-h PCA analgesic consumption prediction.
An analysis of the occurrence frequency of each attribute and its level in the C4.5 bagging trees identified the 10 most informative attributes for the prediction of 72-h total analgesic consumption (i.e., a continuous dose and PCA dose) and the prediction of PCA analgesic consumption only.
Among the most informative attributes, we found HP1α and CTCF downstream of the alternative exon in relation to inclusion events; and AGO1, H3K36me3 and RNAPII in relation to skipping events.
To select the most informative attributes for the classification, we used a combination of feature selection methods (Methods) (Additional file 1: Table S6).
We first extract the sequence-order-independent informative attributes from web page request sequences; these attributes represent a web user's activeness, pages of interest, and the breadth and intensity of their interest.
Compared with the negative logarithms of the average p-values of the informative continuous-dose and PCA-dose attributes (i.e., 152.4 and 11.4), these results suggest that these informative attributes were not identified by chance.
