Hi @Max (Nixtla)
Just sharing the questions here:
Error measure questions
• What's a reliable error measure for evaluating models at the item level, given our data is extremely intermittent?
• Why is the gap between poor and good methods small when calculating error measure at the item level, but becomes larger when we aggregate forecast to a higher level?
• How should we account for in-stock and out-of-stock periods in forecast accuracy calculation at the item level?
• What's the most effective approach for clustering time series data that includes both slow and fast-moving items to identify suitable forecasting methods?
• What are the key features we should consider to classify time series, allowing for the use of distinct models for each category in the presence of intermittency?