Papers in this category often use datasets of 100K+ users to predict psychological traits or engagement.

If your interest is in the algorithm itself applied to this scale:

: Researchers frequently use Random Forest models to analyze large-scale CSV/XLSX exports of Facebook data to predict user attributes like age, gender, or political leaning.

: Unlike "black box" deep learning, RF allows for "feature importance" analysis, showing exactly which Facebook metrics (e.g., shares vs. comments) are the strongest predictors.