Movie Recommendation System using Collaborative Filtering

This repository contains Python code for building a movie recommendation system using collaborative filtering techniques. Below is a breakdown of the files and functionalities included:

1. Installation
2. Code Description
3. Author
4. License

Installation

1. Clone this repository:

   git clone https://github.com/sadegh15khedry/MovieRecommendationSystem.git
   cd Movie-Recommendation-System-Using-Collaborative-Filtering
   

2. Install the required libraries using the environment.yml file using conda:

   conda env create -f environment.yml
   

3. Download the movieLens datasets (movies.csv, tags.csv, ratings.csv) and update the path to them in the code.

4. Run the recommendation_system.ipynb notebook to generate movie recommendations.

Code Description

1. Data Loading and Preprocessing

• Load the datasets (movies.csv, tags.csv, ratings.csv) using pandas.
• Select relevant columns (tag_df, rating_df, movie_df) for further analysis.
• Perform exploratory data analysis (EDA) to understand data shapes, missing values, duplicates, and basic statistics.

2. Data Aggregation

• Merge rating_df and movie_df on movieId to create a combined DataFrame (df).
• Aggregate ratings to find movies with more than 100 ratings (agg_df).
• Merge df with agg_df to filter out less popular movies (df_gt100).

3. User-Movie Matrix

• Create a user-movie matrix (user_movie_matrix) using pivot_table, where rows represent users, columns represent movies, and values represent ratings.

4. Normalization and Similarity Calculation

• Normalize user_movie_matrix (matrix_norm) by subtracting the mean rating of each user.
• Calculate cosine similarities (user_similarity and movie_similarity) based on matrix_norm to find similar users and movies.

5. Recommendation Process

• Select a user (picked_userId) and set up variables (number_of_simlar_users, user_similarity_threshold).
• Find similar users (similar_users) based on a similarity threshold.
• Identify movies watched by the selected user (picked_user_watched) and similar users (similar_users_movies).
• Calculate item scores (item_score) based on weighted sums of ratings from similar users.
• Sort and print top recommended movies (ranked_item_score) based on their scores.

Results and Outputs

• The script outputs top recommended movies for a selected user (picked_userId) based on collaborative filtering.
• Evaluation metrics (e.g., precision, recall) and visualizations (e.g., heatmap of similarity matrices) can be added for performance analysis.

Further Improvements

• Implement evaluation metrics to quantify the performance of the recommendation system.
• Optimize code efficiency for larger datasets and real-time recommendations.
• Incorporate content-based filtering or hybrid approaches for improved recommendation accuracy.

Author

• Sadegh Khedry

License

This project is licensed under the Apache-2.0 License - see the LICENSE.md file for details.