AWS Forecast Inference optimization

Question

Current behavior :

• target series of 1320 data points, which has an hourly frequency, until a specific date (let's say: 2020-01-30 23:00:00)
• related time series from the same date and includes information for the next hour, (e.g. 2020-01-31 00:00:00)
• a Prophet predictor with a horizon of prediction of 1 hour

The approach is to obtain the data for the last recent hour, and include it to the target series (and related time series), in order to forecast the following period. Here is the steps :

• Save the recent updated target time series to S3
• Import the data into Forecast using forecast.create_dataset_import_job
• Delete previous Forecast using forecast.delete_forecast
• Create a new Forecast (without retraining the predictor) using forecast.create_forecast

But the main issue is that the forecast creation takes up to 30 minutes, so is there a way to make the inference faster? or change the approach to optimize the time for doing so?