Purdue Dining Court Food Waste Tracker

Inspiration

We had noticed significant wastage of food in Purdue Dining courts, especially when food was served that students did not enjoy as much. Having a strong interest in using machine learning to combat climate change, we chose to build a neural network that can be used by Purdue's dining staff to minimize food wastage and maximize student's preferences.

What it does

We built a neural network that takes in a string of ingredients or other methods used to make a food item. This is in the case that a dining court decides to introduce a new meal item. The network checks against previous foods that have similar ingredients or recipes and suggests what the approximate student rating would be, the approximate trays needed to be made by the staff and an estimate of the food that will be wasted. We also decided to build a website that displays the rating, approximate trays required for the food, and the weight of the food wasted if a staff member inputs ingredients. On the student view of the website, there is a form that takes in the meal number and the rating of that meal to develop the data for our neural network.

How we built it

We built the neural network on Google Colab using the pytorch framework. We also attempted to build a website using html, python and javascript.

Challenges we ran into

Some challenges we ran into was that it was difficult to find a large dataset for this project and create a neural network with strong layers to compute the text classification with high accuracy.

Accomplishments that we're proud of

We managed to built a neural network from scratch that can take in the ingredients and predict a student rating. We also managed to manipulate online datasets to produce sample data that can help train the algorithm

What we learned

We learned more about dataset manipulation using pandas and creating neural networks for natural language processing using pytorch. We learned how to do basic natural language processing for multiclass text classification.