Time series prediction with linear regression

[luccalb]

User Story

1. As a RTDIP user
2. I want to predict the future development of periodical, time series data, given past values
3. So that I can plan ahead

Additional context

• This component should work with time series data in a fixed interval (every second, every 10 minutes, 1h, ...)
• Let the point in time be t=0 and data is available from t=-100, then train a regression model on this timespan
• The component receive a time in the future (e.g. t=50), up to this point the data should be predicted using the model
• Something like this can be done using scikits LinearRegression https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LinearRegression.html which will try to fit a function to the observed data

Acceptance Criteria

• A new component is implemented
• Component input
  • past data
  • the name of the column with the data that should be predicted
  • point in time in the future, up to which the data will be generated (or the number of data points to predict in the same interval)
  • (optional) a prediction method to use
• Component output
  • past data + future predictions
  • (for all other columns except the predicted one, the values can be empty)
• The component will analyze past, time series, data and train a regression model
• The component will generate new data up to a given point of time in the future, using this model

Definition of Done

• Test cases have been created and are running successfully
• Documentation for the new component was added
• Github Actions are running without errors