Kishan Manani- Backtesting and error metrics for modern time series forecasting | PyData London 2024

[@fmind-dev]

Great talk. I will recommend it to newcomers in AI/ML forecasting.

[@prasvasu4217]

Also just like absolute or root mean square errors, does it make sense to systematically check for forecast bias between specific cohorts or sub-cohorts while back-testing in retail? Why or why-not?

[@prasvasu4217]

Why is it that we want symmetry to over/under-forecasting in retail; don't we want our model to penalize under forecasting more? Generally revenue or profit $ from lost sales is more than reduction in revenue from overstocking correct? That is why big retail eg. fast fashion have always operated in manufacturing to scale push style.