Zee_Recommender_System

• This project builds a personalized movie recommender system using the MovieLens dataset to enhance user experience by suggesting movies based on user ratings and preferences.

• The system implements multiple approaches:

  • Collaborative Filtering (User-based & Item-based)

  • Pearson Correlation

  • Cosine Similarity (with KNN)

  • Matrix Factorization (Surprise Library / CMFRec)

📂 Dataset

• ratings.dat: UserID, MovieID, Rating, Timestamp

• users.dat: User demographics (Gender, Age, Occupation, Zip-code)

• movies.dat: Movie Title & Genres

🚀 Features

1. Data Preprocessing

• Merged ratings, movies, and users into a single dataframe
• Extracted Release Year from titles and handled missing values
• Aggregated average ratings and number of ratings per movie

2. Exploratory Data Analysis (EDA)

• Visualized rating distribution by genre, age, and popularity
• Analyzed average ratings vs. number of ratings

3. Recommendation Engines

• Item-based (Pearson Correlation): Recommend 5 similar movies
• Cosine Similarity (KNN): Generate user-item similarity matrices
• Matrix Factorization (d=4): RMSE & MAPE evaluated; embeddings visualized

4. User-based Collaborative Filtering

• Find top 10 most similar users
• Recommend top 10 movies based on weighted ratings

🛠 Tech Stack

• Python, Pandas, NumPy, Matplotlib, Seaborn

• Scikit-learn (Cosine Similarity & KNN)

• Surprise / CMFRec (Matrix Factorization)

📈 Evaluation

• Root Mean Squared Error (RMSE)

• Mean Absolute Percentage Error (MAPE)

• Visualization of embeddings (d=2 & d=4)

📌 Future Work

• Deploy as a Flask/Streamlit Web App

• Incorporate content-based filtering (genres & metadata)

• Implement hybrid recommendation models