Delivering Locally Sourced Nutritious Food to Indian Households

---

NOTE TO SELF:

• tell sanchita about using softmax
• ask about rice + wheat. Include? Not include?
• revisit the terminology "bin-packing". Is this really a bin-packing problem? The baskets are designed with continuous (not discrete) amounts of each crop.

for sanchita

• Figure out units for production file (same as consumption?)
• Per household minimum (pulse vs millet)
• Check replicability + correctness of ode (after it's finished)

for reuben

• Make algebraic forulation of both problems (see if they agree with each other)
• Add legend to cluster map (in general make it readable)
• Seed clustering so it's reproducible
• Get everything ready for the big model

Replication instructions

Rough outline

1. Dependencies + environment
   1. Python packages: sklearn, matplotlib, numpy, pandas
   2. CPLEX runtimes: Basket Design, Distribution
2. Preference Clustering
   1. Run clustering.py to assign clusters and preferenes
   2. Visualize data with explore_consumption.ipynb
3. Basket Design
   1. CPLEX walkthrough
   2. postprocess walkthrough
4. Distribution

Data explanation

1. Consumption
2. Travel Times
3. Production

References

License

TODO:

• /data
  • clean up unused parts of 2011.xlsx
• /basket_design
  • connect to output of /clustering
  • add postprocessing
• / distribution
  • clean up unused parts
  • change to real travel distance times (tt_ij from google maps)