Each line in the contribution analysis result table is one factor item.

The contribution score is an overall measurement that combines the scale of the actual value and the deviation.

For example: The metric "count of Traffic Size" on iOS as platform dimension is 0.00, while the expected value should be 62.99. So the deviation is 100% down. Considering the pair of actual value and expected value as (0.00, 62.99) and (0.00, 6299), the deviations are both 100% down. However, the contribution score of (0.00, 6299) is larger than (0.00, 62.99), because the value scale of 6299 is larger.

For any anomaly on the target metric, does the following:

1. Calculates the correlation between the target metric for the anomaly and all other metrics.
2. Selects the top correlated metrics.
3. Gets the dimension data of each correlated metric.
4. Calculates the change of dimension data on both the anomaly day and normal days.
5. Lists the top N items as the most potential drivers for the anomaly.