Implement data infrastructure for a modelling framework
Opened this issue · 1 comments
Currently, AequilibraE does have a structure to represent a full model. This includes attributes such as:
- Time period definitions
- scenario data
- Metadata for matrix cores
- Based on the OT PMTURI concept?
- Socio-economic data (population data is being done in Tradesman)
- 4-step model parameters?
- Model iteration?
- Further details on the results database
- Results other than link flows
I reckon its about 50/50 where the benefits of processing origin, destination zone based variables as 2d vectors (computation) outweighs the benefit of just having stacked data with a multi index in pandas (stored as parquet). PMTURI is cool but I still think it locks us into storing many things as matrices that probably just dont belong. I really like xarray. It just generalises the PMTURI concept as "dimensions" and "coordinates" as discussed in xarray docs and allows for label based ndimensional workings. In the model I made I chose explicitly to have fiteen minute as by most disaggregate temporal unit (96). Time periods are just a coordinate of that. I also dont have the "direction" dimension because I have not estimated destination choice with PA/AP format. But yeah all good I think you guys are doing great. Just some two cents from me.