Kuiz berkaitan panduan · Sederhana Tahap

Kuiz Pemarkahan Ketua AI

Menilai cara pemarkahan petunjuk menggunakan hasil sejarah, mengapa ketidakseimbangan kelas mempengaruhi penilaian, dan mengapa penentukuran dan pemantauan drift penting.

Laluan panduan yang berkaitanPemarkahan Utama AI
soalan 1 daripada 8

Bagaimanakah pemarkahan pendahuluan AI berbeza terutamanya daripada pemarkahan berasaskan peraturan tradisional?

Pratonton setiap soalan8 soalan

Jawapan dan penjelasan muncul semasa anda mengambil kuiz.

  1. Bagaimanakah pemarkahan pendahuluan AI berbeza terutamanya daripada pemarkahan berasaskan peraturan tradisional?
    • Ia mempelajari pemberat daripada data penukaran sejarah dan bukannya menggunakan mata tetap
    • Ia hanya mengira e-mel dibuka
    • Ia tidak memerlukan data sama sekali
    • Ia mengabaikan saiz syarikat dan jawatan pekerjaan
  2. Tugas pembelajaran mesin manakah yang paling sesuai menerangkan meramalkan sama ada petunjuk akan bertukar?
    • Pembahagian imej
    • Pengelasan binari
    • Pengecaman pertuturan
    • Pembelajaran pengukuhan
  3. Mengapakah ketepatan biasa merupakan metrik yang lemah untuk menilai model pemarkahan petunjuk?
    • Accuracy directly measures calibration
    • Accuracy tells sales representatives why a lead will convert
    • Converters may be rare, so a model can appear accurate by predicting “no conversion” for nearly every lead
    • Accuracy automatically adjusts decision thresholds
  4. According to HubSpot’s lead-scoring documentation, which data can a custom score use?
    • Only a person’s name
    • Hanya masa hari
    • Unlabeled audio waveforms
    • Record actions or properties selected for the scoring model
  5. What can a lead-scoring model return for each prospect?
    • A score or estimated conversion probability, which still needs evaluation
    • A probability that is automatically calibrated by definition
    • A guaranteed sales outcome
    • A legal determination of creditworthiness
  6. What does feature importance summarize in a tree-based lead-scoring model?
    • That a feature causes conversion
    • How influential a feature is in the model’s decisions under that importance method
    • That a feature is legally required
    • The future sales quota
  7. What does calibration mean for a lead-conversion score?
    • Every lead receives the same score
    • The model’s accuracy is always 100 percent
    • Predicted probabilities match observed conversion frequencies over comparable cases
    • Scores are sorted alphabetically by company
  8. Why should a lead-scoring model be monitored after deployment?
    • The original threshold can never become outdated
    • Monitoring makes labels unnecessary
    • A model score permanently captures market behavior
    • Lead behavior and the meaning of historical patterns can drift