M365 Copilot had assigned that exercise, but as I mentioned earlier, I skipped it. I didn’t quite grasp what it was asking for—I just did what I thought made sense and moved on to number 6…
After finishing number 6, I asked it exactly what it wanted me to do, since I hadn’t understood the previous task. It explained something like this: you need to generate normal data and insert an anomaly. The dataset would consist mostly of normal values, with very few abnormal ones. It provided this pseudocode to illustrate the logic:

if random < error_chance:
    generate unusual value
else:
    generate normal value