Table 1.
Performance of machine learning classifiers for predicting gun ownership and gender.
| Measure and data set (purpose) | Best classifier: accuracy | Best classifier: F1-score | |
| Gun ownership |
|
|
|
|
|
Linked survey or Twitter sample (cross-validation average) | 0.72 | 0.69 |
|
|
Holdout from linked survey or Twitter sample (internal validation) | 0.71 | 0.70 |
|
|
Gun conversation sample (external validation) | 0.65 | 0.64 |
| Gender |
|
|
|
|
|
Gender sample (training) | 0.90 | 0.87 |
|
|
Holdout from gender sample (internal validation) | 0.90 | 0.88 |
|
|
Linked survey or Twitter sample (external validation) | 0.74 | 0.73 |