Any plans to add predictions from regression models?

[kha84]

Hello there! Imagine you have a data source, whose data metrics are subject of oscillation. The nature of that oscillation is a combination of seasonality (e.g. more in winter, less in summer), trend-following (e.g. growing year to year), well-predicted spikes (e.g. more data on weekend, but less during public holidays) and whatever other measily-meatbags-related reasons data is depending on :)

I found it pretty useful to fit a regression model to historical metric values, so I can just monitor that actual latest metric values are within the trusted interval of the prediction from the model. This approach reduces false alarms, a usual companion of any predefined thresholds for data quality metrics.

Will be great to see the same in Checkita.

[gabb1er]

Hello! We are glad to announce, that in latest release we have added two types of trend metrics:

• linear regression metric - fits simple linear regression model over historical results treating referenceDate as X-vector and metric results as true Y-vector. Metric predicts result for current referenceDate.
• ARIMA metric - trains ARIMA model over historical metric results. Metrics forecasts the next value in time-series.

For more information, see docs.

We have more plans regarding time-series forecasting, so I will keep this issue open for now. In meantime, we will be glad to have some feedback regarding new features.

[kha84]

Great! Thanks a lot!