Solving hunger in Africa

A modeled approach to solve this chronic issue

0
0
meals were wasted in Europe meals were needed in Africa

in the 0 seconds since you opened this website

Why are there people who are still starving?

Exactly what we wondered too! So we got down to work.

Despite recent technological improvements, there is still a large share of starving people on this globe. The aim of this project is to propose a way to end starvation in Africa, one of the continents most affected by this hunger issue. To do so, we’ll devise an appropriate daily nutritional input with the goal of determining a set of food products. To this end, a dataset of the Food and Agricultural Organization of the United Nations (FAO) was used as a primary resource of information. It contains material about agricultural production amounts, producer prices and a lot of other useful details to determine which countries should provide which food items, and is therefore at the heart of this project.

Our final goal:
Coming up with a possibility on how to redistribute food while minimizing the overall effort and costs. Ever since the dawn of civilization, the African continent was riddled by chronic food shortages and famines. With growing affluence in recent years, the European continent on the other hand has experienced resource abundance. As a consequence, society's awareness for the dilemma of food waste has significantly increased. However, specific proposals on solutions are rare and often too optimistic. So why not use Data Science to solve it?



A history of food distribution

How much does each continent have?

As a starting point, looking at the total food supply every year in each country already enables a solid overview about the dynamics. More specifically, Africa and Europe’s historical food situation was thoroughly analyzed. Based on past values, data was predicted for the upcoming year 2020 in order to be as close to reality as possible. The following maps show the food that was supplied by each continent for every year since 1960.


In Africa, there is a clear disparity between North African countries and their sub-Saharan counterparts, which started to take form in the 70's. Currently, all of the former ones command a supply of more than 3000 kcal per person per day except for Sudan. To put this into context: A man at the age of 20 is expected to have at least 3000 kcal per day to safely avoid malnutrition. Simply judging by the change in color over the years, it can be stated that overall, there already was some improvement in most nations, especially after the year 2000. Some of the states were not listed in the FAO dataset and are consequently blank.


As expected, most of the European countries show a stable food supply situation, with values being more homogeneous than on the African continent. Interestingly, Belgium, Austria, and Italy are provided with the highest amount per capita. Impacts of historical events can be seen when simulating a time series using the sliders. For example, a sharp drop in Eastern Europe can be observed in the years following the dissolution of the Soviet Union. Furthermore, a correlation with a nation’s Gross Domestic Product (GDP) can be assumed.

Again, a few countries were not included in the dataset (mostly recenly established nations) and are left blank with no values assigned to them.


How much does each continent need?

In order to deduce a country’s demand, it would seem natural to just look at the population size multiplied by an average person’s daily need. However, this methodology would not account for different demographic compositions . According to this source, there are discrepancies between varying age groups and their respective nutritional requirements. The plot clearly shows how the highest caloric need is addressed by males and females in their 20s and 30s.


By multiplying this data with the same demographic age-range information, a much preciser number for national food demand is obtained. This takes situations into account where a nation’s population composition differs substantially to others. For example, during war periods the male population would significantly be reduced. Additionally, the age-sex-pyramid often shows higher shares of young people and babies, who evidently need less than an adult person.


How much is Africa lacking and how much can Europe help?

Subtracting demand from the supply, deficits for the case of negative results could be determined. Move the sliders to get insight on the food surplus' evolution over time!


Various events can be put into context here, for example, the famine in Ethiopia in the ’80s, where Ethiopia figured among the countries with the highest deficit per capita. Looking at the more recent years, this animation already suggests that only by smart redistribution, Africa could sustain its own food demand . However, having the capabilities and know-how to efficiently set up a food aid operation is harder than it seems. European countries, on the other hand, have a lot more experience in this field, and they are expected to have an even higher amount of excess food, making it easier to provide for this whole operation. Ever since 2001, more than half of the examined countries show a net surplus. As of 2019, all of the countries in red were determined to be either war-riddled or politically fragile. Exceptions to the rule are Namibia and Eswatini, both of which boast a relatively high GDP per capita (ranked 10th and 11th for the African continent, respectively). Thus, the only explanation would be inequality amongst the population or insufficient distribution of available resources.
Surprisingly, all European countries boast an overall surplus in the food supply, with the exception being Albania which was slightly deficient during the first few years of our investigation (‘60s). Another significant occurrence can be detected during the aftermath of Yugoslavia’s disintegration in 1992, where ex-member states like Croatia, Bosnia & Herzegovina and former Macedonia’s nourishment status deteriorated drastically, causing a temporary shortage of nutrition. Most of the other countries, however, have constantly shown a surplus of around 1000 kcal per day and per person, meaning a third of European food does not end up being consumed - in other words: food waste.


Europe to the rescue

The surplus in 2020: A global view

Determining recipient countries should be easy, one might think. Just consider the red ones, right? Yet what about unequal wealth distribution? A better-off person eating more than the calculated daily average means another person gets less. For this reason, a threshold of 300 calories in surplus was introduced to be on the safe side. This also entails more countries are being targeted to receive help. Let us now list all the previously determined values for 2020 on a map. This yields the following:


The observations mainly coincide with what was expected. The results seem to be similar to the food supply situation, only minor changes for some countries can be detected. The degree of food deficit of Chad in Northern-Central Africa, for example, was mitigated by the fact that its population is very young, compared to other African nations. These small differences, however, are crucial when evaluating which countries are actually in need. Using a map for visualization further allows us to observe that mainly landlocked countries in Central Africa show a large food deficit. Considering that countries like Ethiopia, which are known to be prone to suffer from famines were also determined to be lacking in this analysis, confirming that the methodology and the datasets that were used can be assumed to be accurate. With the previously defined surplus threshold of 300 kcal/capita/day , the red and orange countries in the map to the left will be considered to receive food aid.


Which countries can help

Let’s determine now which European countries would be most suitable for redistribution. Belgium boasts the highest surplus per capita (around 1350 kcal/capita/day ). Just to put this into perspective: a Belgian person’s equivalent of food waste could help one person in every single deficient African country each to reach a healthy amount of calory intake. When redistributing food, most larger European countries have enough of a surplus to solely solve the hunger issue in Africa. However, if only one country starts to donate its entire surplus, the costs are assumed to be greater than if multiple countries decided to send a smaller amount each. On the other hand, the amount of European countries should be narrowed down in order to facilitate calculations and enable a proper analysis.



In this case, Germany, France, Italy, Spain, and the United Kingdom were chosen as they occupy the top-right corner of the plot, therefore they have the highest GDPs and food supply per capita. Moreover, the markers' size is considered as it is an indicator of the total food surplus, that in turn takes into account of the population. Even though Russia has the greatest total surplus, it was not considered as its GDP is much smaller than the one of other European countries. Moreover, its vast extent and remote nature makes efficient endeavors of collecting remaining food unlikely. Spain was preferred over Poland due to more useful geographical position as well as better economic performance.

Let us now have a look at how to evaluate a specific redistribution.


Our solution

The perfect diet

Now, the exchange of calories is known. However, in what form should it be delivered? To reach a concrete and tangible recommendation on how to solve this issue, this is an essential question. For a well-balanced diet, a share of 55% carbohydrates, 25% proteins and 20% fat is recommended. Secondly, products of varying food groups should be included.


To this end, a ranking system was introduced to analyze all listed food items based on their previously defined composition suitability. In doing so, the list of products to be considered could be reduced to 13 items to be used for further analysis. The food categories consequently considered are the following ones:
  • Beef Products
  • Cereal Grains and Pasta
  • Fruits and Fruit Juices
  • Poultry Products
  • Legumes and Legume Products
  • Vegetables and Vegetable Products
Selecting a data point in the plot on the right will allow you to see the product's name and its composition of the three macronutrients projected onto the corresponding axis.


The distribution plan

For every European countries out of the five selected, domestic prices for specific items are found to be fairly different. In this context, the natural way to go is to redistribute food in the most economic-advantageous way as the goal of the project is not to penalize European countries for their food waste but rather create the condition for a win-win scenario. Consequently, a cost minimization is carried out using optimization methods (if interested, check out the appendix). Furthermore, a linearly increasing cost function with respect to the GDP is implemented yielding a higher cost if only one country was considered while also prefering to take from richer countries.


This chord plot provides a general overview about how food could be redistributed with minimal expenditures. When hovering over a country's circle, its associated food flows will be highlighted. By selecting a specific country in the drop menu, you can reduce complexity and assess where this particular nation's food is meant to be allocated. Note that the total amount of food sent/received corresponds to the extension of a country's circular arc. Germany and France were found to contribute the most, with each providing slightly more than 100,000 tons. This is mainly due to their higher GDP. In contrapposition, Italy and Spain are found to contribute the least and this again confirms the goodness of our hypothesis. It is trivial to see that Ethiopia would claim the highest share of food aid (1.3*105 tons) of all examined countries. Considering that a regular Ethiopian person is estimated to lack 125 kcal every day, the intermediate result is quite promising as it indicates that the extent of population is correctly taken into account.


In a final step, we move to a detailed view of what amount of each specific product should be shipped to every individual African country. By clicking on an individual country, its food to be received will be displayed. The products Beef, Meat, Chicken, Oat and Tomatoes are required the most in all countries in need. This can be explained by the fact that - compared to the other items - they have a lower cost per kcal of energy. Meat is used that much due to its high content in valuable proteins and fat, which other food items are mostly lacking. The respective shares of food aid are similar for all African countries, as their deficit was assumed to be comprised of the same lack of carbohydrates, fat and protein.

Conclusion

Most importantly, it was demonstrated that the food being wasted in Europe exceeds by far what would actually be required to end hunger in Africa. Most of the major European nations could actually solve the problem on their own by smart redistribution. However, a combined effort of multiple nations resulted in a cost-effective scenario in which the most suitable European countries could actually solve the African hunger issue. When considering specific food items to be shipped, a diverse selection of products is determined, ensuring a well-balanced diet which enables an active lifestyle.
One of the most impactful parameters of this model is the threshold of food surplus, where countries located below it are considered to be in need. Changing this value by just a few calories greatly alters the model's results. For future references, this variable should therefore be meticulously assessed and implemented. To do so, each country's wealth distribution should be evaluated to detect the degree of food consumption imbalances within the population.
Lastly, if considering African starvation problem by an economic point of view, the results of our cost analysis and its minimization could be further used to conduct a feasibility study. Governments and International Agencies could estimate explicit costs on food-aid shipments in order to allocate a specific budget for this purpose.

Appendix

Datasets used

  • FAO dataset
    This dataset contains the following sub-dataset, which are used:
    • Food balance: This section contains food supplies for every country in the world. As the documentation reports, a region supply is defined as: “Production + imports - exports + changes in stocks (decrease or increase)”. We will use this database to analyze the amount of kcal/person/day for each African *and* European state in the food categories we are interested in.
    • Prices: We used FAO dataset to obtain prices of the food items analysed in our diet. In this context, FAO did not contain all the prices needed.
    • GDP: GDP was retrieved and used in order to obtain an overview on how much each European country should contribute to the cause.

  • United Nations Population Dataset
    This dataset was used to determine the male and female population for Europe and Africa from 1950 to 2020 with a granularity of 5 years. We will interpolate to obtain data with 1 year frequency.

  • Agricultural Research Service (ARS) Nutrition Facts Database
    Used to build the optimal diet. This dataset contains nutritional information for raw products. The information reported are per 100 g of servings.

  • European Commission dataset
    This dataset is used to get domestic food prices for items whose prices were not included in the FAO dataset. In the end, the majority of the prices came from this dataset.

  • Additional
    Estimated calories needed: we scraped this website to obtain information on the average calories needed for the African population.

Methods

Recurrent Neural Networks (RNN)

This widely used machine learning method was used to predict data up until the year 2020, as FAO only provides information until 2013. The RNN were preferred over a simple linear regression because the latter does not take into account intermittent outliers like periods of war or polticial instability. Additionally, the mostly exponential growth of food supply was not considered in this method. Also, giving the short avaibility of data (only about 50 datapoints for each country), using the feeback provided by the RNN improved the accuracy of the model exponentially. The RNN was fitted using all the data from the past, but an additional time window method was implemented. This method uses as datapoint to predict the next year supply not only the past history (already embedded in the model), but also a time span of the last 10 years datapoints, in order to fit better the time-series behavior of the recent history. Tensorflow's tutorial was used and adapted to fit this model.

Convex Optimization

Two minimizations were carried out.
The first tackled the issue of minimizing the amount of food that has to be delivered by the five chosen european contries. While minimizing, the demand constraints had to be respected as well as the constraints on food availability of each European country. The objective function to be minimized was a quadratic non-negative weighted sum of food [kcal/year]. More spefically, the weights were designed to take into account both the GDP of a country and the food availability [kcal/year]. The modelling choice is justified by the fact that a rich country with a large surplus should contribute more than a relatively poor country with less possibilities. The problem we want to model is a Quadratic Program with Linear Constraints (QP) . The choice of a quadratic objective function is due the fact that we will need to evenely distribute the resources and this means that weights will have to increase quadratically with the amount of food given away. It makes sense to say that the more a country gives away of its surplus, the less the same country should contribute further if other countries didn't countribute at all yet.
The second optimization concerned the composition of the diet, which is one of the key steps of our analysis. We decided to compute the amount of product by minimizing the actual costs of shipments that every European country had to meet. In order to model the problem we needed to define an objective function which in this case was the non-negative weighted sum of products' cost. The problem we want to model is a Linear Program.