🎓 Personalized Learning Recommender System

A powerful and interactive course recommendation system built with Streamlit, combining both traditional machine-learning algorithms and deep-learning models to help learners discover the best courses tailored to their personal preferences.

---

📌 Overview

The explosion of online education platforms has empowered millions—but it also overwhelms learners with too many choices. This project solves that problem by providing personalized course recommendations using:

• Content-based filtering
• Collaborative filtering
• Clustering
• Deep learning models

The system offers an intuitive Streamlit interface where users can select courses they’ve taken, choose a model, tune hyperparameters, train real-time, and instantly view recommendations.

---

🚀 Features

• Interactive UI: Built with Streamlit + AgGrid for fast filtering and selection.
• 8 Recommendation Algorithms .
• Real-time Model Training: Train Neural Networks or KNN directly in the browser.
• Hyperparameter Tuning: Sliders for k-value, epochs, clusters, etc.
• Cold-Start Support: Automatically builds a user profile for new users.
• Embeddings-Based Models: Neural networks + regression/classification using learned embeddings.

---

🧠 Models Implemented

1. Content-Based Filtering

• Cosine similarity over processed course descriptions.

2. User Profile Model

• Builds a weighted user vector based on genre affinity.

3. K-Means Clustering

• Recommends courses popular within the user's cluster.

4. PCA + Clustering

• Dimensionality reduction for improved clustering performance.

5. KNN (User-Based / Item-Based)

• k-Nearest Neighbors collaborative filtering.

6. Neural Network Recommender

• Custom Keras model learning low-dimensional embeddings.

7. Regression with Embeddings

• Linear regression using NN-learned embeddings.

8. Classification with Embeddings

• Random Forest classifier using learned embeddings.

(Note: NMF model removed as per latest project update.)

---

🛠️ Installation

Follow the steps to run the project locally:

1. Clone the Repository

git clone https://github.com/pradeep-mahat0/course-recommender-system.git
cd course-recommender-system

2. Create a Virtual Environment (Optional)

Windows

python -m venv venv
venv\Scripts\activate

macOS / Linux

python3 -m venv venv
source venv/bin/activate

3. Install Dependencies

pip install -r requirements.txt

4. Run the Application

streamlit run recommender_app.py

---

📖 Usage Guide

1. Launch the app — the course dataset loads into an interactive grid.
2. Select courses you’ve previously taken or liked.
3. Choose a model from the sidebar.
4. Tune hyperparameters using sliders.
5. Train the model (only for training-based models like Neural Network).
6. Click Recommend New Courses to see personalized suggestions.

---

📂 Project Structure

├── recommender_app.py      # Main Streamlit UI
├── backend.py              # Model controller and logic layer
├── nn_model.py             # Keras-based embedding model
├── helper.py               # Utility functions for processing & clustering
├── recommend.py            # Cluster recommendation logic
├── data/                   # CSV datasets
│   ├── ratings.csv
│   ├── course_processed.csv
│   └── ...
└── requirements.txt

---

🔮 Future Enhancements

• Database Integration (PostgreSQL/MySQL) for scalable user data.
• Hybrid Filtering combining content + collaborative scores.
• Explainable Recommendations with SHAP, attention maps, or feature attributions.
• API Layer to allow mobile app or external service integration.

---

👨‍💻 Created by Pradeep Mahato

If you like this project, don’t forget to ⭐ the repository!