Table 4 Two dimensions results comparison \(|1-scaledfactor|\) : the TDGGD between (\(X'\xrightarrow {m}Y'\)) and (\(X'\xrightarrow {f}\tilde{Y'}\)); the predictions between (\(X \rightarrow Y\)) with (\(X'\xrightarrow {f}\tilde{Y'}\)) on the TDGGD and ShipGen.

From: A tabular data generation framework guided by downstream tasks optimization

Target column

\(Y'\)

\(\tilde{Y'}\)

Difference

Ratio

EI+HI+AI

EI+HI

Difference

Ratio

Cw

0.3418

0.0219

– 0.3199

– 93.59%

0.0219

0.3958

0.3739

94.47%

SA1

0.2430

0.3621

0.1191

49.01%

0.3621

0.5202

0.1581

30.39%

SA2

0.4589

0.7742

0.3153

68.71%

0.7742

1.2183

0.4441

36.45%

Vol1

0.2273

0.2239

-0.0034

-1.50%

0.2239

0.5620

0.3381

60.16%

Vol2

0.2031

0.2499

0.0468

23.04%

0.2499

0.3930

0.1431

36.41%

MB

0.0434

0.2017

0.1583

364.75%

0.2017

0.2262

0.0245

10.83%

GC

0.1397

0.9516

0.8119

581.17%

0.9516

1.9513

0.9997

51.23%

Mean

0.2367

0.3979

0.1611

141.66%

0.3979

0.7524

0.3545

45.71%

  1. The best results are marked as bold values in X data and bolditalics values in Y data.
  2. The entire results on entire methods can be found as Supplementary Tables S7 and S8.