Recommendation system: evaluation index [offline evaluation index: RMSE (root mean square error), AUC, precision, recall, F1] [online evaluation: A/B test] [generally required response time <0.5s]

I. Evaluation process

2. Evaluation Dimensions

1, user dimension

2, platform dimension

3. Engineering dimension

III. Offline evaluation

1. Offline evaluation metrics

1.1 RMSE (Root Mean Square Error)

1.2 Precision, Recall, F1

Precision and recall are two complementary metrics, and one should not be emphasized.To visually observe these two metrics at the same time, we can introduce the F-value, which is equal to the harmonic mean of precision and recall.

1.3 Logarithmic Loss Function

1.4 AUC

1.5 MAP (mean precision mean)

2. Offline evaluation method

Test set cannot be earlier than trainingCollection

IV. Online assessment

---

---

---

References:
Recommendation system series from 0 to 1—8_Common indicators and open source framework