citlacom
commented
3 years ago Hi @thiloyes - I did the refactoring needed to persist and extract the models prediction date from a metadata file. Also I have refactored the predictions performance report script to more granular functions so is easier to extend and maintain. Let me know if you need anything else to calculate the predictions performance significance. Thanks!
In order to allow computation of the models predictions accuracy from all unseen data that start at the models predictions creation, we need to store this date in a metadata file due the fact that every time that a user clones the repository, the date of predictions are set to the clone date and not the predictions creation.
Additionally we need to update the modelsPredictionsPerformanceReport script to calculate model production days from metadata date instead of file information date.
Once this is resolved @thiloyes will be able to validate if the models accuracy that have persisted during 70 days or more of unseen data could confirm that planets cycles are really correlated and that the great predictions results is not just a coincidence. For that purpose, some probabilities calculations will be performed to determine the likelihood that just for coincidence a machine learning training phase resulted in similar accuracy just by chance and based on that determine machine learning models results significance.
There are more traders in Trading View "Astrology, Cycles & Gann Trading" and in our "Financial Astrology Research Group" Telegram chat that have confirmed that in their research have concluded that financial astrology works really well.