Figure 3.

Ranks of the top 10 variables for the prediction of length of hospital stay (LOS). Variables were ranked using a classifier-specific evaluator based on machine learning algorithms. Each variable was ordered according to their mean ranks. The lower rank represents more contributions to the prediction of LOS. For example, tourniquet time was ranked 1st, 2nd, and 3rd in GBDT, RFC, and XGB, respectively. TQT, tourniquet time; FRT, femoral resection thickness; TCS, tibia component size; IBL, intraoperative blood loss; FCS, femoral component size; IT, insert thickness.