Table 3 Normalized permutation importance (sum to 1.0) of 6 opinion-related factors in their corresponding random forest regression models and their \(R^2\) values. The 15 explanatory models are generated based on the results of 10,000 simulation instances, each with the corresponding testbed. The result shows that, among the 6 factors, the average of the public’s initial opinion values (\(O_1\)) dominates the rest in terms of the permutation importance

Testbed 1

0.8082

0.8453

0.0622

0.0232

0.0107

0.0091

0.0496

Testbed 2

0.8491

0.8477

0.0024

0.0485

0.0027

0.0007

0.0980

Testbed 3

0.8490

0.7972

0.0304

0.0307

0.0137

0.0070

0.1209

Testbed 4

0.7969

0.8970

0.0024

0.0308

0.0034

0.0014

0.0650

Testbed 5

0.8831

0.8699

0.0030

0.0495

0.0043

0.0007

0.0726

Testbed 6

0.9573

0.9046

0.0006

0.0312

0.0100

0.0002

0.0535

Testbed 7

0.8233

0.9351

0.0406

0.0067

0.0043

0.0016

0.0118

Testbed 8

0.7931

0.7609

0.0230

0.0198

0.0029

0.0023

0.1912

Testbed 9

0.8720

0.7834

0.0338

0.0595

0.0289

0.0081

0.0863

Testbed 10

0.9898

0.9429

0.0087

0.0028

0.0010

0.0001

0.0445

Testbed 11

0.9786

0.9665

0.0029

0.0085

0.0011

0.0002

0.0207

Testbed 12

0.8548

0.8751

0.0761

0.0095

0.0086

0.0074

0.0233

Testbed 13

0.8483

0.8215

0.0047

0.0277

0.0061

0.0022

0.1378

Testbed 14

0.9670

0.8581

0.0003

0.0535

0.0056

0.0003

0.0821

Testbed 15

0.7698

0.6967

0.0285

0.1091

0.0350

0.0143

0.1163