I am new to SciDB and I have a specific use case in mind for my project. I am thinking of doing in-database analytics for feature engineering for the models I want to build.
Here is my use case for my current project. I am essentially building a model would would have featured derived from the primary transactional rows like the following –
average meal prep time over the past 1 day, average meal prep time over the past 1 week, average delivery time for order at that time of the day over the past 1 month, average weather conditions at that time, number of items in the order, order amount quantiles etc.
As orders get inserted into the system, I am imagining that I can create derived tables that would compute these features for me.
Once the features are derived, I would like to normalize and scale the features and would apply certain matrix transformations before pulling them out and building some models.
Is this something I can do today with SciDB? I also want to find out if a Time-Series database will be better suited to this since a lot of the features are about computing averages and quantiles over a period of time.