Compare results of the gradient boosting model outlined in pictures below. Describe and explain differences in model performance. Identify the role and impact of predictor variables in the CART and gradient boosting models.

Model Summary: Model error measures Name Learn Test Average LogLikelihood (Negative) 0.09206 0.12288 ROC (Area Under Curve) 0.96758 0.88413 Variance of ROC (Area Under Curve) 0.00013 0.00175 Lower Confidence Limit ROC 0.94506 0.80225 Upper Confidence Limit ROC 0.99011 0.96600 Lift 8.96104 6.84211 K-S Stat 0.84819 0.62195 Misclass Rate Overall (Raw) 0.03422 0.03283 Balanced Error Rate (Simple Average over classes) 0.08304 0.22951 Class. Accuracy (Baseline threshold) 0.88896 0.84848 Fraction Data Used after influence trimming 0.37379 naVariable Importance Variable Score ACT 100.00 AT 77.52 MKVALT 72.03 AP 61.24 SALE 57.78 LCT 52.26 CAPX 49.40 AM 40.84 DLC 34.681 | | | |IIIIII OPTVOL 20.641 /IIISummary for 192 Trees - Gains Chart - ROC, Sample: Full sample, Target class: 1 The Fox. Rate 1 1 30107030405060/080510 False Ros . RateConfusion Matrix - Test Actual Total Percent Predicted Classes Class Class Correct 0 N=313 N =83 0 377 81.96% 309 68 19 78.95% 4 15 Total: 396 Average: 80.46% Overall % 81.82% Correct: Specificity 81.96% Sensitivity/ 78.95% Recall Precision 18.07% F1 statistic 29.41%