How did you choose the threshold of 0.6?

The threshold of 0.6 was chosen based on the analysis of the model's metrics and business requirements. Usually, the choice of threshold depends on a compromise between precision and recall.

Process of threshold selection:

1. Plotting the ROC curve or Precision-Recall curve on the validation set.
2. Analyzing metric values at different thresholds.
3. Choosing a threshold that provides an acceptable balance between Type I and Type II errors, considering the specifics of the task.

For example, if minimizing false positives is more important, the threshold may be higher; if not missing positive cases is more critical, it may be lower. In our case, 0.6 showed an optimal balance, confirmed by testing on real data.