Recommender-System-Lab-Assignment

Lab: Building a Simple Recommender System (Based on Chapter 2)

Objectives

By the end of this lab, you will be able to: - Understand the difference between Content-Based and Collaborative Filtering. - Build a simple recommender system using Python and scikit-learn. - Evaluate recommendations using similarity measures.

---

Prerequisites

• Python 3.x\

• Install required libraries:

  pip install pandas scikit-learn

---

Dataset

We will use a small Movie Dataset with the following structure:

The data is attached in the repository

Exercise Steps

1. Load the Dataset

import pandas as pd

# Load dataset
df = pd.read_csv("movies.csv")
print(df.head())

---

2. Create a User-Item Matrix

user_item_matrix = df.pivot_table(index="user_id", columns="movie_title", values="rating")
print(user_item_matrix)

---

3. Apply Collaborative Filtering (Cosine Similarity)

from sklearn.metrics.pairwise import cosine_similarity

# Fill NaN with 0 for similarity calculation
matrix_filled = user_item_matrix.fillna(0)

# Compute similarity between users
similarity = cosine_similarity(matrix_filled)
print("User Similarity Matrix:\n", similarity)

---

4. Make Recommendations

# Example: Recommend for user 1 based on most similar user
import numpy as np

user_index = 0  # user_id = 1
similar_users = similarity[user_index]

# Find the most similar user (excluding self)
most_similar_user = np.argsort(similar_users)[-2]

# Get movies rated by most similar user
recommended_movies = user_item_matrix.iloc[most_similar_user].dropna().index.tolist()
print(f"Recommended movies for User 1: {recommended_movies}")

---

5. Experiment

• Try Content-Based Filtering using movie genres.\
• Compare results from different approaches.

---

Deliverables

• Submit your movies.csv dataset.\
• Upload your Python notebook or .py script with code and results.\
• Write a short reflection on:
  • Which method (Content-Based vs Collaborative) worked better?
  • What challenges did you face (e.g., missing data, new users/items)?
• Push your code to the GITHUB repositery and atatch the link for submission

---

Extension (Optional)

• Implement a Hybrid Recommender (combine content + collaborative).\
• Use a larger dataset like MovieLens.