I have always seen data modeling as a framework for the visualization of problems and their solutions, thereby making them easier to understand and solve. The issue of food insecurity is one which is close to my heart. I live in a nation of 200 million where, to this day, eating 3 square meals a day is a privilege. Food insecurity is a real life or death issue where i am from. Unfortunately, my country doesn't have an affinity for data collection and descriptive statistics. My hope with this project is to gain capital and experience in order to tackle world hunger issues. I believe that we should strive to make access to quality and nutritious food at dirt cheap prices. Food should be a human right, and access to food is a major determinant of susceptibility to bribery and corruption especially with regards to the electoral process in developing nations where absolute poverty is rife.
Families in remote and isolated indigenous communities in northern cnada frequently lack access to affordable nutritious foods, particularly perishables such as fruits, vegetables and meats, due to limited food selections, high food prices and poor quality of fresh produce. Expensive transport costs and difficult logistics (e.g. airfreight charges, and uncertainty of travel on winter roads, where they exist, or air travel subject to weather conditions), high poverty rates and a continuing decline in the use of traditional foods result in few healthy food choices.
What it does
Nunafood is an ML algorithm which aims to determine the amount of discount necessary to achieve balance sheet equilibrium between personnel, transportation and storage costs and the amount and price of excess produce(excess or otherwise) pre-wholesale, thereby reducing waste at the governmental and food production levels.
How I built it
I first built this project using microsoft excel and the spyder data analysis tool from the anaconda toolbox. I had to canvas for the relevant data, determine factors which influenced the success of this venture and build two(2) datasets of my own.
The first dataset has columns:community ,population ,%of_population ,personnel_costs_as%of_pop ,recommended_personnel_costs_individual ,no_of_personnel ,storage_cost ,personnel_cost ,total_kg_of_food_allocated_as_a_percentage_of_populationfor_60,000_kg_of_food ,freightfare_per_kg_from_winnipeg_to_nunavut communities ,freightfare_per_region_based_off_kg_of_food_allocated.
The second dataset(fresh_produce) has columns:product_type_beef, price_per_one_kg_beef, location_beef produce_lifetime_days_beef, product_type_dairy price_per_litre_kg ,location_dairy ,produce_lifetime_days_dairy ,product_type_fruit, price_per_kg_fruit, location_fruit ,produce_lifetime_days_frui,t product_type_grain, price_per_bushel_grain ,location_grain ,produce_lifetime_days_grain product_type_vegetables ,price_per_kg_vegetables ,location_vegetables ,produce_lifetime_days_vegetables.
We estimate that 60,000 kg is the lower limit of any produce to be enough to feed the currently 35K+ people in the nunavut region.
- I built a method(collectProduce) to collect the price of each produce by kg and multiply it by our total expected kg(60,000) in order to get the price of each produce at a scale of 60000KG
- I built a second method(getProducePriceMinOverhead) to subtract the total overhead costs from the price of the produce in order to get our rough breakeven produce price numbers.
I will follow up with more development but due to the time constraints of submitting this, i must stop here.
Challenges I ran into
This process was ardous, finding relevant data to build the dataset into my python script was quite difficult, i had to draw from a myriad of sources. After building out the dataset, deciding which factors were relevant was quite the task.
Accomplishments that I'm proud of
The 2 goals of this project were:
1- To determine the size of discount per unit of a produce and the cumulative number of produce required to break even on transportation, personnel and storage costs when sending food to Northern Canada.
2- To build a viability model to determine the chances that a good can actually be discounted effectively while still shelling out a stipend per unit of produce.
Building out compelling datasets like the two i use here. It was so difficult to find relevant information and this made it almost impossible. I am still not sure how useful all of this is, but that is for you to determine. I do believe that with the right data, these datasets could transform the way the canadian government subsidizes food going into the north. I can see it reducing the negative balance on the governments sheets and account for overhead while slashing prices based on the percentage of produce going to waste at the farm production level
What I learned
I learned that some of my weaknesses are lifelong battles. Stuff like motivation and existential implications of my success cannot be the primary drivers of my success. I need to feel the urgency every single night and ensure to check for changes to submission schedules and adjust my development schedule accordingly. I think i have failed at it this time. I submit this just to challenge myself to be better.
What's next for Nunafood forecaster
I plan to keep improving this algorithm and dataset even after my submission and i will till i feel happy and comfortable in my decision to work on this project.