To improve accuracy, the base models provided by TimesDB can be fine-tuned in the following ways:

  • Using Datasets: Existing models can be fine-tuned using data from one or more datasets. This can be done by referencing the dataset IDs or tags.
  • Using SeriesIDs: Models can be fine-tuned using one or more input series. Each prediction accepts an input series and generates a predicted series.
  • Using Series: Models can be fine-tuned using actual data points, represented as (timestamp, value) pairs.

The resulting fine-tuned model will have a unique model Id. Additionally, tags can be assigned to them for better organization and categorization.

Model tags are unique across all models. This means if a tag is applied to a model, the same tag will be removed from any other model it was previously assigned to.