Implement optimization/scenarios framework

[lmnoel]

[jkoschinsky]

See Jamie's code.

Questions to address with what-if scenarios:

Adding or removing records and recomputing indicators (per capita spending and access):

• compared to current allocations, how do indicators change if new allocation is added in a particular location
• … or current one is removed?

Compared to current spending, how does per capita spending and access change under different funding mechanisms:

• equal distribution per capita
• prioritization of high need

Optimization questions to maximize citywide access and per capita spending:

• where to put new 500k for specific service ?
• where to re-allocate existing spending to avoid over- or undersupply?
• where are gaps in service provision where new investments would be needed (for what service type)?

[jkoschinsky]

FYI: Google's optimization tools: https://developers.google.com/optimization/

[jkoschinsky]

Notes:

• simulated annealing: very slow

• potential solution: turn objective function into a constraint: optimize for an access score bin (e.g. increase in good access bin)

• use access score bins instead of continuous score from 0-100

• save scenarios + cache, so they can pre-run and then then take to meetings

• data in browser: WebGL — transfer data from server to browser

[jkoschinsky]

Nick Mader's slides, incl. simulated annealing, FYI: https://github.com/nsmader/CANOPY/tree/master/write-ups/presentation