Table 5 Evaluation of machine learning estimates (at different levels of τ against human coded sample (n=1,000).

From: Policy agendas of the American state legislatures

τ

Precision

Recall

F1

0.9

0.59

0.23

0.31

0.8

0.55

0.25

0.32

0.7

0.52

0.30

0.36

0.6

0.49

0.32

0.37

0.5

0.45

0.41

0.39*

0.4

0.42

0.49

0.41

0.3

0.41

0.54

0.44

0.2

0.37

0.61

0.43

0.1

0.27

0.69

0.38

  1. *Default level of τ used to generate estimates.
Back to article page