Skip to main content
. 2019 Nov 20;7(4):e15601. doi: 10.2196/15601

Table 2.

Prediction performance using the 9 selected important features.

Performance measure Logistic regression (ridge), n (%) Logistic regression (LASSOa), n (%) Random forests, n (%)
Accuracy (low volatility class; n=694)

Subsample 1 510 (73.5) 513 (73.9) 461 (67.4)

Subsample 2 518 (74.6) 516 (74.4) 475 (68.4)

Subsample 3 511 (73.6) 518 (74.6) 474 (68.3)

Subsample 4 511 (73.6) 506 (72.9) 454 (65.4)

Subsample 5 504 (72.6) 506 (72.9) 455 (65.6)

Consolidated 510 (73.5) 515 (74.2) 476 (68.6)
Accuracy (high volatility class; n=185)

Subsample 1 114 (61.6) 122 (65.9) 119 (64.3)

Subsample 2 116 (62.7) 117 (63.2) 129 (69.7)

Subsample 3 114 (61.6) 115 (62.2) 121 (65.4)

Subsample 4 118 (63.8) 121 (65.4) 124 (67.0)

Subsample 5 120 (64.9) 119 (64.3) 123 (66.5)

Consolidated 115 (62.2) 121 (65.4) 125 (67.6)
Overall accuracy (n=879)

Subsample 1 624 (71.0) 635 (72.2) 587 (66.8)

Subsample 2 634 (72.1) 633 (72.0) 604 (68.7)

Subsample 3 625 (71.1) 633 (72.0) 595 (65.8)

Subsample 4 629 (71.6) 627 (71.3) 578 (65.8)

Subsample 5 624 (71.0) 625 (71.1) 578 (65.8)

Consolidated 625 (71.1) 636 (72.4) 601 (68.4)

aLASSO: least absolute shrinkage and selection operator.