# Suppress warnings for a clean notebook
import warnings; warnings.filterwarnings("ignore")
import numpy as np
np.random.seed(42)
 

# Data setup
import pandas as pd, zipfile, requests, io
url = 'https://files.grouplens.org/datasets/movielens/ml-latest-small.zip'
r = requests.get(url)
z = zipfile.ZipFile(io.BytesIO(r.content))
df = pd.read_csv(z.open('ml-latest-small/ratings.csv'))
print(df.shape)
print(df.head(3))

# Check basic info about our data
df.info()
 

# Let us look at the unique users and movies
num_users = df['userId'].nunique()
num_movies = df['movieId'].nunique()
print(f"There are {num_users} unique users and {num_movies} unique movies.")

# Let us find the average rating given by each user
user_means = df.groupby('userId')['rating'].mean()
print(user_means.head())

# Let us explore ratings for a specific movie
movie_stats = df.groupby('movieId')['rating'].agg(['count', 'mean']).sort_values(by='count', ascending=False)
print(movie_stats.head(5))

# Create a user-movie rating matrix
rating_matrix = df.pivot_table(index='userId', columns='movieId', values='rating')
print(rating_matrix.shape)
print(rating_matrix.head(3))

# Visualize the distribution of ratings
import matplotlib.pyplot as plt
import seaborn as sns
sns.histplot(df['rating'], bins=10, kde=True)
plt.xlabel('Rating Value')
plt.title('Distribution of Movie Ratings')
plt.show()

# Let us try user similarity: compute cosine similarity
from sklearn.metrics.pairwise import cosine_similarity
user_similarity = cosine_similarity(rating_matrix.fillna(0))
print(user_similarity.shape)
print(user_similarity[:3, :3])

# Pick a random user to recommend movies for
user_id = df['userId'].sample().values[0]
print(f"Selected user id: {user_id}")

# Find the most similar users to our target user
target_idx = rating_matrix.index.get_loc(user_id)
sim_scores = user_similarity[target_idx]
similar_users = rating_matrix.index[sim_scores.argsort()[::-1][1:6]]
print(f"Most similar users: {list(similar_users)}")

# Gather movies liked by similar users but not yet rated by our user
user_movies = set(df[df['userId'] == user_id]['movieId'])
sim_movies = df[df['userId'].isin(similar_users) & ~df['movieId'].isin(user_movies)]
top_movies = sim_movies.groupby('movieId')['rating'].mean().sort_values(ascending=False).head(5)
print("Recommended movies:", list(top_movies.index))

# Show a mini project: build for a different user
another_user = int(input("Enter a user id to get movie recommendations (pick a number from your earlier list): "))
target_idx = rating_matrix.index.get_loc(another_user)
sim_scores = user_similarity[target_idx]
similar_users = rating_matrix.index[sim_scores.argsort()[::-1][1:6]]
user_movies = set(df[df['userId'] == another_user]['movieId'])
sim_movies = df[df['userId'].isin(similar_users) & ~df['movieId'].isin(user_movies)]
top_movies = sim_movies.groupby('movieId')['rating'].mean().sort_values(ascending=False).head(5)
print("Recommendations for your user:", list(top_movies.index))

# Optional: Map movie ids back to titles for better user experience
movies_url = 'https://files.grouplens.org/datasets/movielens/ml-latest-small.zip'
r = requests.get(movies_url)
z = zipfile.ZipFile(io.BytesIO(r.content))
movies_df = pd.read_csv(z.open('ml-latest-small/movies.csv'))
movie_lookup = dict(zip(movies_df['movieId'], movies_df['title']))
recommended_titles = [movie_lookup.get(mid, "Unknown") for mid in top_movies.index]
print("Recommended movie titles:", recommended_titles)

# Advanced: Try item based collaborative filtering
item_similarity = cosine_similarity(rating_matrix.fillna(0).T)
print(item_similarity.shape)