The aim of this project is to construct a Neural Collaborative Filtering recommendation system and train it on Yelp's dataset to generate personalized restaurant recommendations.

Key Features

• Data Preprocessing: Comprehensive data handling to organize and split the data effectively for training, validation, and testing purposes.
• Hybrid Recommendation Model: An ensemble approach combining Generalized Matrix Factorization (GMF) and Multi-Layer Perceptron (MLP) components to create a robust model capable of understanding complex user-item interactions through both linear and non-linear computations.
• Hyperparameter Tuning with Ray Tune: Employing Ray Tune for adaptive hyperparameter optimization, enhancing model performance through Population-Based Training (PBT).
• Evaluation and Inference: Evaluation of the model's performance on unseen data using RMSE, followed by a practical application scenario demonstrating the system's utility.

References

• The dataset utilized in this project is sourced from Kaggle: Yelp Dataset. Note that the data files are too large to be included directly in the repository, so you will need to download them separately as described below.
• The ensemble approach of the recommendation system developed in this project is largely based on the insights and methodologies presented in the paper Neural Collaborative Filtering by He, Liao, Zhang, Nie, Hu, and Chua (2017).

Getting Started

To dive into this project, clone the repository to your local machine or environment where you have Python (3.8 or newer) and the necessary hardware support (preferably a GPU). The project is contained within a Jupyter notebook, offering an interactive platform to explore and extend the work presented.

Installation

1. Clone the repository to your local machine or development environment.
2. Download the Yelp Dataset from Kaggle. Ensure you place the folder named yelp-dataset within your project directory, containing yelp_academic_dataset_business.json and yelp_academic_dataset_review.json.
3. Install the required dependencies using pip install -r requirements.txt.
4. Launch the Jupyter notebook to walk through the project from data processing to making recommendations.