Understanding the backtesting mode
When you add a dataset for backtest purposes, Amazon Lookout for Metrics uses it to analyze your data over multiple intervals to identify patterns and learn about expected variations. The backtest mode allows your detector to learn about your data before it starts processing continuous data. When you provide historical data, Amazon Lookout for Metrics uses 70% to learn from it and 30% to run a backtest.
The amount of data Amazon Lookout for Metrics can use to learn from historical data depends on the interval you configured when you added the dataset. The service will only look for up to 2,500 previous intervals that must fall within a certain timeframe:
- For 5-minute intervals, the previous intervals have to lie within the last 3 months.
- For 10-minute intervals, this timeframe will be 6 months.
- For 1-hour intervals, this timeframe will be 3 years.
- For 1-day intervals, this timeframe will be 5 years.
If you do not provide...