🍽️ Recipe Recommendation System & Sentiment Analysis

University of San Diego — AAI 510 (Team 8 Final Project)

---

🧠 Project Overview

This project builds an intelligent culinary recommendation platform that predicts user preferences and analyzes recipe reviews.
It merges collaborative filtering and sentiment analysis (NLP) to deliver personalized and meaningful recipe suggestions.

🎯 Goal: Enhance user engagement by combining recommendation systems with opinion mining.

---

🎯 Objectives

• Recommend recipes using both user history and ingredient similarity.
• Perform sentiment analysis on user reviews to understand preferences.

---

🧩 Dataset

Food.com Dataset — contains over 500,000 recipes and 1.4 million user reviews.

Each entry includes:

• Recipe ID, Name, Ingredients, and Tags
• User Reviews and Ratings (1–5)
• Review Text for sentiment analysis

Source: Kaggle — Food.com Recipes and User Interactions

---

⚙️ Methodology

1️⃣ Data Preprocessing

• Cleaned and normalized text data (stopword removal, lemmatization).
• Built user–item matrices for collaborative filtering.
• Tokenized ingredient lists for semantic similarity modeling.

2️⃣ Recommendation System

Collaborative Filtering (SVD):

• Generated latent factors using Singular Value Decomposition (SVD).
• Calculated cosine similarity between user and item vectors.
• Predicted unseen ratings based on top-N similar items.

Content-Based Filtering (Word2Vec):

• Used Word2Vec embeddings to capture ingredient similarity.
• Recommended similar recipes for first-time (cold-start) users.

3️⃣ Sentiment Analysis

• Classified review text polarity using traditional and neural models.
• Models used:
  • Logistic Regression
  • Naïve Bayes
  • Bi-LSTM (for contextual understanding)
• Evaluated using Accuracy, Precision, Recall, and F1-score.

4️⃣ Evaluation

• RMSE for rating prediction accuracy.
• Precision@K / Recall@K for recommender performance.
• F1-score for sentiment classification.

---

🧰 Tech Stack

Category	Tools / Libraries
Programming	Python 3.9+
Data Handling	pandas, NumPy
NLP	NLTK, spaCy, Word2Vec
ML / DL	scikit-learn, TensorFlow / Keras
Recommender	Surprise (SVD), Cosine Similarity
Visualization	Matplotlib, Seaborn
Environment	Jupyter Notebook / Google Colab

---

Key Takeaways:

• The hybrid SVD + Word2Vec system improves personalization.
• Sentiment-aware filtering enhances overall recommendation quality.

---

🚀 How to Run

1. Clone the Repository

git clone https://github.com/seema469/AAI_510_Team8_FinalProject.git
cd AAI_510_Team8_FinalProject