Table 1 Model parameter of RF model predicting change to Annalena Baerbock (M1) and Armin Laschet (M2) based on pre-debate and RTR data.

From: How to convince in a televised debate: the application of machine learning to analyze why viewers changed their winner perception during the 2021 German chancellor discussion

 

Change to Baerbock (M1)

Change to Laschet (M2)

Sample size

4613

4613

Frequency of class labels

No change: 3522, Change: 1091

No change: 4188, Change: 425

Number of trees

1250

1250

Average no. of terminal nodes

248.78

107.939

No. of variables tried at each split

250

250

Total no. of variables

314

314

Analysis

Random

Random

Forest

Forest

Classification

Classification

Number of random split points

30

30

(OOB) Normalized Brier score

35.45505

12.84104

(OOB) AUC

93.21869

95.85909

(OOB) Error rate

0.12378062,

0.08006814,

0.26489459

0.04010405,

0.01886342,

0.24941176