Research Article

Perceptron Ranking Using Interval Labels with Ramp Loss for Online Ordinal Regression

Table 5

Sensitivity estimation of PRIL, PA-I, and proposed PRIL-RAMP algorithm of three datasets under 75% noise data.

DatasetRoundsMethodMAERMSESpearman’s correlation coefficientDiscard sampleDiscard rate (%)Time (s)

AbaloneT = 1400PA-I-75%0.000.206
PRIL-75%0.000.190
PRIL-RAMP-75%61.140.230
T = 2800PA-I-75%0.000.368
PRIL-75%0.000.350
PRIL-RAMP-75%64.780.528
T = 4100PA-I-75%0.000.624
PRIL-75%0.000.575
PRIL-RAMP-75%64.780.681

Parkinsons-uprdsT = 2000PA-I-75%0.000.273
PRIL-75%0.000.288
PRIL-RAMP-75%72.800.352
T = 4000PA-I-75%80.000.587
PRIL-75%0.000.365
PRIL-RAMP-75%871.950.703
T = 5800PA-I-75%0.000.899
PRIL-75%0.000.928
PRIL-RAMP-75%68.561.036
Real estate valuationT = 140PA-I-75%0.000.124
PRIL-75%0.000.180
PRIL-RAMP-75%65.000.144
T = 280PA-I-75%30.000.268
PRIL-75%0.000.388
PRIL-RAMP-75%66.780.270
T = 400PA-I-75%0.000.595
PRIL-75%0.000.516
PRIL-RAMP-75%75.500.402