Inspiration:
Over the years, fast food and eating out have become the norm. As the typical adult becomes busier and the general idea of the nuclear family shifts, individuals are presented with more obstacles with putting a healthy meal on the table. So, our team looked for a way to utilize the boom of sales from grocery delivery and pickup to encourage more home cooking without overwhelming consumers.
Product Information:
We built a functional web application to showcase software that could possibly be used for consumers using an online grocery store. The idea is when a user inputs their grocery list, or from an online grocery store standpoint- when the user adds an item to their online cart, the application provides a list of recipes the user can make using the items in their cart. The list of recipes primarily consists of things they already plan on purchasing and some recipes that only require a few more additional ingredients. This is determined by the machine learning algorithm of Cosine Similarity. The user also has the option to filter out the recommended recipes based on allergies and other dietary restrictions.
Tools utilized to build the product:
We prominently used Jupyter Notebook and Python to write the code, along with Flask to execute the code into a web application. We also used HTML/CSS for the web application functionality.
Challenges we ran into:
Machine learning was not the strongest suit for all team members, so we had trouble getting started with writing the code and implementing a proper machine learning algorithm. We explored a few machine learning algorithms such as clustering but decidedly ended up going with the cosine similarity. We initially used a massive dataset (~200,000 recipes), and the average runtime to deploy our application is ~15 minutes. We decided to cut down our dataset to 10,000 recipes for a feasible runtime (less than a minute). Additionally, integrating the front-end and back-end was also challenging as we ran into several debugging issues with Flask.
Accomplishments we’re proud of:
We are proud of being able to put out a code that functions well despite not everyone in the team having the same background and experience with data science and machine learning. Throughout the coding process, we each contributed and gained knowledge from others’ contributions that will help us with other projects we build in the future.
For the future:
Our team plans to work on the front end heavily. Currently, the code executes, but the web application is not ideal for consumers to use. We plan to use a larger dataset, deploy the code to the cloud, and use the engine to run it. We also plan to add healthier and eco-friendly recipes that support sustainability and add more filters to make the application suitable for more individuals. Additional future direction involves point systems for consumers that try the recipe recommendations.
Log in or sign up for Devpost to join the conversation.