Course Recommender System

This project is a content-based recommendation system designed to recommend Coursera courses based on user search queries and search history. The system provides two types of recommendations:

1. Similar Courses: Shows courses similar to the currently searched course based on course content and features.
2. Personalized Recommendations: Offers personalized course recommendations based on the user's search history.

Project Workflow

The project workflow is as follows:

1. Data Preprocessing:

   • Clean and preprocess the Coursera dataset, one-hot encoding categorical variables and translating course text data to English.
   • Vectorize course summaries using the CountVectorizer for similarity calculations.

2. Exploratory Data Analysis (EDA):

   • Visualize data distributions and proportions of different attributes like course duration, difficulty, and certificate types.

3. Course Search and Recommendation:

   • Uses cosine similarity to find and recommend courses based on a user's current search or past searches.
   • Updates a user vector to track and personalize recommendations based on search history.

Running the Project

Clone this repository and navigate to the project directory.

Run the Streamlit application with the following command:

streamlit run app.py

Once the app is running, it will open in a web browser. Use the sidebar to explore recommendations or view visualizations.

Data Preprocessing

One-Hot Encoding: The dataset contains categorical features (organization, course time, and difficulty) which are one-hot encoded.

Text Processing: The text fields (course_summary, course_description, course_skills) are cleaned, tokenized, and stemmed. The course_summary field is further processed to create a text vector.

Vectorization: Using CountVectorizer, the processed text data is transformed into a numerical format for similarity computations.

Recommendation Logic

Similar Courses For each searched course, a similarity index is calculated using cosine similarity on the vectorized text features. The system then retrieves the top similar courses based on this similarity score.

Personalized Recommendations A user vector is initialized and updated with each search to track the user's course preferences. Each course search creates a list of similar courses, forming a retrieval list. Recommendations are refined by calculating the dot product of the user vector with courses in the retrieval list, resulting in personalized suggestions.