Algorithm · Google · Hard
Part 1: Robust percentage RMSE You are given a list of revenue records. Each record is a triple (market, observed, forecast), where market is a string label, observed is the actual revenue, and forecast is the predicted revenue. For any collection of records, define the percentage root-mean-square error (pRMSE) as $$ \text{pRMSE} = \sqrt{\frac{1}{\sum_i w_i} \sum_i w_i \left( \frac{forecast_i}{observed_i} - 1 \right)^2} $$ where the sum runs over records that pass the…
Checking your access…