I think option D looks promising - using feature crosses between the location and car type features could help the model learn the nuanced relationships in each city. I'd want to experiment with that approach.
Okay, let me think this through. The key things I'm considering are the user-specific nature of the data and the need for easy schema maintenance. Cloud Spanner could be a good option, as it's a fully managed relational database with automatic schema updates. But Cloud SQL is also a solid choice, and might be a simpler transition from the existing MySQL setup.
Roslyn
4 months agoFelice
4 months agoRenea
4 months agoAlline
4 months agoJess
5 months agoPauline
5 months agoHoney
5 months agoCarolann
5 months agoFrederica
5 months agoLavelle
5 months ago