---

RecStudy: A Comprehensive Study of Recommender Systems

Welcome to RecStudy, an introductory journey for beginners in recommender systems. This project covers various techniques, from basic collaborative filtering to deep learning algorithms and advanced graph-based and knowledge graph methods.

The primary reference for this project is Professor Yu Fangren's book, "Hands-on Learning of Recommender Systems: Algorithm Implementation Based on PyTorch". A significant portion of the code is inspired by the content of the book and the author's code repository. Special thanks to Professor Yu Fangren for his invaluable contributions!

To facilitate learning for beginners, all code is thoroughly commented, well-organized, and ready to run without any adjustments. You can directly execute the code and see the final results. If you encounter any issues, feel free to raise them in the Issues section!

Below is the directory structure and a detailed description of each chapter.

---

Project Structure

RecStudy/
├─ dataset/                        # Datasets used in the project
│  ├─ ml-100k                      # MovieLens 100K dataset
│  ├─ ml-100k-original             # Original MovieLens 100K dataset
│  └─ ml-latest-small              # MovieLens Latest Small dataset
├─ Part1.BasicRecAlgorithms/       # Basic recommender system algorithms
├─ Part2.AdvancedRecAlgorithms/    # Advanced deep learning-based algorithms
├─ Part3.GraphRecAlgorithms/       # Graph-based recommender algorithms
├─ Part4.KGRecAlgorithms/          # Knowledge graph recommender algorithms
├─ Part5.ConstructRecSys/          # Recommender system construction techniques
├─ Part6.EvaRecSys/                # Evaluation methods for recommender systems
└─ utils/                          # Utility scripts

---

Chapters Overview

Part 1: Basic Recommender Systems

This chapter introduces foundational algorithms and techniques for recommender systems, including matrix factorization, collaborative filtering, and similarity-based methods.

File	Description
ALS_PyTorch.py	Alternating Least Squares (ALS) implemented in PyTorch.
ALS_tradition.py	ALS implemented in NumPy.
basicSim.py	Basic similarity computation for user-item pairs.
evaluate.py	Evaluation metrics for recommender systems.
FM.py	Factorization Machines (FM) implementation.
FM_embbeding_style.py	FM with embedding style implementation.
furtherSim.py	Extensions of similarity computation.
itemCF_01label.py	Item-based collaborative filtering with binary labels.
LR.py	Logistic Regression-based recommendation.
POLY2.py	Polynomial-based recommendation model.
SVD.ipynb	Singular Value Decomposition (SVD) for recommendations.
userCF_01label.py	User-based collaborative filtering with binary labels.
userItemCF_15label.ipynb	Hybrid user-item collaborative filtering with detailed labels.

---

Part 2: Advanced Recommender Systems

This chapter explores deep learning-based approaches, such as neural collaborative filtering, embedding methods, and sequence-based models.

File	Description
ALS_MLP.ipynb	Neural Collaborative Filtering (NCF) with MLP.
Embedding_CNN.ipynb	Embedding-based recommendation using CNNs.
Embedding_MLP.ipynb	Embedding-based recommendation using MLP.
FNN_plus.ipynb	Feedforward Neural Networks with additional features.
RNN_rec.ipynb	Sequential recommendation using RNNs.
RNN_rec_ALS.ipynb	Combining RNNs with ALS for hybrid recommendations.

---

Part 3: Graph-Based Recommender Systems

In this chapter, we introduce graph neural networks (GNNs) and other graph-based techniques for recommendations. (Files in this directory to be updated.)

File	Description

---

Part 4: Knowledge Graph Recommender Systems

This chapter focuses on leveraging knowledge graphs to enhance recommendation quality. (Files in this directory to be updated.)

File	Description

---

Part 5: Constructing Recommender Systems

This section delves into building end-to-end recommender systems, including system design and implementation. (Files in this directory to be updated.)

File	Description

---

Part 6: Evaluating Recommender Systems

This chapter discusses evaluation techniques and metrics for recommender systems, including accuracy and diversity measures. (Files in this directory to be updated.)

File	Description

---

Dataset

The project uses several datasets, primarily the MovieLens dataset, with different versions organized as follows:

• ml-100k: Preprocessed MovieLens 100K dataset.
• ml-100k-original: Original MovieLens 100K dataset.
• ml-latest-small: Latest small version of MovieLens.

---

Utilities

Additional scripts and helper functions are stored in the utils/ directory. These include data preprocessing and utility scripts to facilitate the project.

File	Description
dataloader.py	General data loader for recommender systems.
dataloader4ml100kIndexs.py	Data loader for MovieLens 100K with indexed features.
dataloader4ml100kOneHot.py	Data loader for MovieLens 100K with one-hot encoding.

---

Getting Started

To start exploring the code and running experiments:

1. Clone the repository:

   git clone https://github.com/Embraccce/RecStudy.git
   cd RecStudy

2. Install dependencies:

   pip install -r requirements.txt

3. Run examples or experiment scripts from respective directories.

---

Contributing

We welcome contributions to improve and expand this repository. Feel free to submit issues or pull requests.

---

License

This project is licensed under the MIT License.

---