With a growing population and degrading environment, better decisions must be made to more sustainably distribute food. According to the Food and Agriculture Organization of the United Nations, ⅓ of food produced for human consumption is lost or wasted.1 Determined to reduce these statistics using data, the team decided to tackle the problem with retailers, because the highest waste per ton occurs at the consumption and distribution stages. After reaching out to
a range of local grocery stores and international retailers, the team realized that data on sales and inventory for food waste is either not available or companies were not willing to share it.
Due to the lack of availabµle food waste data, the team decided to pivot when the opportunity to partner with an international, French retailer on a project to predict demand on promoted products presented itself. By helping the company to more accurately predict demand and understand the impact of promotions in their Italy stores, the team is indirectly achieving their overarching goal of reducing food waste that is caused by overstock items.

The team analyzed two years of sales data from the company’s stores. After cleaning, engineering features, and testing a wide variety machine learning algorithms on the data, the team was able to develop a model that predicts the daily revenue product sales quantity based on features such as promotions, time, and product nomenclature. Despite the pivot from analyzing food waste data to analyzing sales data, we are still achieving our goal in reducing food waste as helping to more accurately predict demand and understand the impact of promotions indirectly reduces food waste by preventing an overstocking of items.
While much of the team had basic experience in Python and data science projects, by the end we had become far more proficient with our data manipulation and Python skills, as well as gaining experience outreaching to industries. Ultimately, our team is intending to continue working with the French retailer in the future, where we will continue engineering features and building models to find the best ways to combat food waste.

https://github.com/AnneSpitz/data-x-fall2018-foodwaste