The main goal of the recipe recommendation system was to develop a user-friendly dashboard that recommended healthy and tasty recipes to college students based on their preferences and nutritional needs. To achieve this goal, I started by cleaning and preprocessing the recipe dataset from Kaggle, which included over 50,000 recipes with directions, ingredients, reviews, and total calories for each recipe.
Next, with help from my team, I incorporated a dataset that focused on the ingredient-level nutrition of these recipes to give a more comprehensive review of the nutritional value of each meal. The team and I used cluster analysis, specifically K-means clustering, to group together recipes that were similar to the user’s preferences. After normalizing the dataset, weights were added based on user inputs, allowing the preferred features to have more influence on the clustering.
Finally, K-nearest neighbors was used to recommend a certain number of recipes based on the user’s preferences and displayed them with the dashboard, along with images and directions for each recipe. The “image” capabilities of the Plotly Dash package was implemented to display the images of each recipe, making it easier for college students to plan their meals and improve their diets.
Skills: Python and Excel