Pedro Capelastegui

At the moment, run_forecast takes candidate trend and seasonality models, and fits all possible combinations. This gets very inefficient as new models are introduced. We should move towards a stepwise...

model_decay - remove redundant parameters

1

Current implementation of model_decay uses 4 parameters. We can remove at least one safely.

In metadata, reformat params_str

1

column params_str in the forecast metadata table currently has the following format: ``` [-4.3e+02 -2.4e-04 5.3e+02] ``` We could make it more useful by adding commas: ``` [-4.3e+02 -2.4e-04 5.3e+02]...

Error: Initial guess outside of bounds

2

We are using the following custom model: ```python model_step_lockdown = forecast_models.get_model_step_date('2020-03-16') forecast_models.model_linear+model_step_lockdown ``` On the initial value check, we get an out of bounds errors: ``` AssertionError: Initial guess outside...

Output index should probably be named 'date' or 'x', depending on the case.

There should be an email address or mailing list for users to submit questions or comment topics that fall out of the scope of github issues.

Add seasonal model - hourly

1

Found some use cases with hourly sampled data - time to implement the hourly seasonal model

Add residuals checks

1

Our prediction intervals sometimes have a trend, suggesting that the residuals for that model have a non-zero null. This points to a sub-optimal fit - we should check for it...

This is not intended. We need to constrain this model to only fit seasonality for periods up to a year. 

in run_forecast(), parameter date_start_actuals can be used to ignore actuals samples before a specific date. Currently, we just filter out data before that date. It would be best to use...