A social-network platform wants to detect and reduce spammy friend requests.

How would you define a spammy friend request and choose discriminative features? 2) Describe how you would build a classification model. 3) If no labeled data exist, what strategies would you use? 4) After deployment, how can the model be leveraged to improve user experience? 5) How would you balance precision and recall and decide on the operating threshold?

Think through labeling strategy, feature engineering, semi-supervised learning, precision/recall business impact, and product feedback loops.

Meta Data Scientist interview question: Detect and Reduce Spammy Friend Requests Effectively. {"blocks": [{"key": "c2927356", "text": "Scenario", "type": "header-two", "depth": 0, "inlineStyleRanges": [], "entityRanges": [], "data": {...