Overordering in restaurants leads to food waste, a lot of food waste. In Iceland, we throw away 5.840 tons of food every year. Food waste has a huge environmental impact that we cannot afford in the midst of climate change. Food waste releases greenhouse gas emissions during farming, processing, shipping, distribution, storage, and disposal. Food waste has a very large carbon footprint, water footprint, and land usage. We are dedicated to improving the wellbeing of our planet. After having personally seen buckets and buckets of food thrown away through a decade worth of experience in the foodservice industry, we knew we had to something.

What it does

GreenBytes reduces food waste by telling restaurants how much food they should be ordering. We have developed a cloud-based solution that helps food retailers reduce food waste and increase profits by breaking down restaurant menus and predicting future food consumption using our newly adapted machine learning algorithms.

Our solution takes daily sales data into account to predict the amount of food necessary for the upcoming days. We use a Recurrent Neural Network (RNN) to predict the amount of each product that will be sold in the upcoming days based on past sales data as well as weather forecasts for the area a restaurant is located in.

How we built it

The weather data that was incorporated into the model were obtained from Veðurstofnun. This data was used with our already existing data from a restaurant in Reykjavík.

All fo the data was cleaned and visualized. The correlation between total sales and each weather variable was assessed and the relevant elements were included as parameters to train the RNN.

After the RNN was trained, predictions were made for a predetermined amount of days into the future. The amount of each ingredient required to fulfill the predicted orders was determined.

To validate the problem two steps were taken.

1) The predictions were compared to the true number of items sold. The average Root Mean Square (RMS) Error for the predictions was 2.5 with an average product number of 6.6

2) The known amount of food that was ordered for the restaurant was compared to the amounts that we predicted would be necessary to fill the order. The amount of waste that was produced for those three days was 648 kg. With the predictions, the amount of food that would be produced was -19 kg, The negative indicates that the model slightly underestimated for some items.

Challenges we ran into

We ran into challenges getting the predictions into a data frame to compare how much food is needed

Accomplishments that we're proud of

We are proud of raising awareness about food waste and working to reduce food waste. We understand that food waste has an impact on our planet and our society, making it a problem worth solving. We are proud to provide a solution that can help restaurants benefit the planet and stay in business. We are living through a complex time where climate change and a pandemic our constantly on our minds. We think of GreenBytes as the contribution we can make to help chip away at the climate change problem while helping restaurant owners stay in business.

At the core of Greenbytes is our belief in our local communities and the individual’s ability to affect environmental change. If we can empower domestic businesses to improve the food supply chain, we can help Iceland serve as an international role model for better consumption and production practices throughout Europe.

Europe throws away 88 million tonnes of food per year which costs €143 billion, reducing this by 20% can save Europe €28.6 billion.

The GreenBytes team has the industry knowledge and business model to be part of the conversation that starts with improving Iceland’s domestic food supply chain. Our solution can help Iceland reach its Paris agreement goals, as well as our domestic goal of becoming carbon neutral by 2040. GreenBytes can help fulfill the following UN Sustainable development goals: Responsible Consumption and Production (12), Climate Action (13), Zero Hunger (2), Gender Equality (5), and Reduced Inequalities (10).

What we learned

Over the past week, we learned about the correlation between weather and food sales. We learned that the climate factors that have the greatest correlation between the total sales and the weather/time data are gusts, max wind speed, precipitation, snow cover, day of the week, and sunlight hours.

We learned that there are many datasets available in Iceland that could aid is in more accurately predicting the amount of food requirements for commercial food retailers. With some more research, we can more accurately predict the quantities of food and have a greater impact on reducing food waste

What's next for GreenBytes

The next steps for GreenBytes will be to further analyze the algorithms, conduct lifecycle assessment analyses, and conduct full-service trials.

Now that we have seen an improvement in our overall algorithm accuracy through the introduction of external factors such as weather, we will seek out other datasets and see if we can further draw connections between food demand and external factors.

We want to better understand the impact we are making in the restaurants we work with. In order to do so, we need to accurately quantify the waste we reduce and analyze what that means in terms of carbon emission reduction, water usage reduction, and land usage reduction.

Lastly, we need to test our full service in restaurants. Testing means having a restaurant input their menu, stock, and distributors and allow our progressive web application to make order suggestions without our interference. Testing the web application will allow us to understand the scalability of our project.

Built With

Share this project: