print('''

- You are given a recommendations.csv file with user ids and you are supposed to recommend atmost 10 songs.
- The training data is provided in train.csv file.
- Your entries in the recommendations.csv file should be such that for each row the first value must be user_id followed by recommended song_ids all seperated by comma.
- Make sure you have atleast one recommendation for each user in recommendations.csv or else your score will be zero
- The recommended songs to a user must be different from what you already have in the training set for the same user.
- A user can have at most 10 recommendations.

# Load the data
train_data = load_train_data("train.csv")
recommendations_data = load_recommendations_data("recommendations.csv")

# Preprocess the training data
user_songs = preprocess_train_data(train_data)

# Build a recommendation model (e.g., collaborative filtering)
model = build_recommendation_model(train_data)

# Generate recommendations for each user in recommendations.csv
with open("recommendations.csv", "w") as recommendations_file:
    for user_id in recommendations_data['user_id']:
        recommendations = generate_recommendations(user_id, model, user_songs)
        recommendations = limit_recommendations(recommendations)
        formatted_recommendations = format_recommendations(user_id, recommendations)
        recommendations_file.write(formatted_recommendations + "\n")
def preprocess_train_data(train_data):
    """
    Preprocess the training data to create a set of songs already listened to by each user.

    Args:
    - train_data (DataFrame): DataFrame containing the training data with columns user_id and song_id.

    Returns:
    - user_songs (dict): Dictionary where keys are user_ids and values are sets of song_ids listened to by each user.
    """
    user_songs = {}
    for index, row in train_data.iterrows():
        user_id = row['user_id']
        song_id = row['song_id']
        if user_id not in user_songs:
            user_songs[user_id] = set()
        user_songs[user_id].add(song_id)
    return user_songs
from surprise import Dataset, Reader, KNNBasic
from surprise.model_selection import train_test_split

def build_recommendation_model(train_data):
    """
    Build a recommendation model using collaborative filtering.

    Args:
    - train_data (DataFrame): DataFrame containing the training data with columns user_id, song_id, and rating.

    Returns:
    - model (surprise.prediction_algorithms.algo_base.AlgoBase): Trained recommendation model.
    """
    # Create a Surprise Dataset object
    reader = Reader(rating_scale=(0, 1))  # Since we only need binary rating (listened or not)
    data = Dataset.load_from_df(train_data[['user_id', 'song_id', 'rating']], reader)

    # Split the data into training and testing sets
    trainset, _ = train_test_split(data, test_size=0.0)  # Use full data for training
    
    # Build the recommendation model using KNNBasic algorithm (or any other algorithm of your choice)
    model = KNNBasic(sim_options={'user_based': False})  # Item-based collaborative filtering
    
    # Train the model
    model.fit(trainset)

    return model
import pandas as pd
from surprise import Dataset, Reader, KNNBasic
from surprise.model_selection import train_test_split

# Load the training data
train_data = pd.read_csv("train.csv")

def preprocess_train_data(train_data):
    """
    Preprocess the training data to create a set of songs already listened to by each user.

    Args:
    - train_data (DataFrame): DataFrame containing the training data with columns user_id and song_id.

    Returns:
    - user_songs (dict): Dictionary where keys are user_ids and values are sets of song_ids listened to by each user.
    """
    user_songs = {}
    for index, row in train_data.iterrows():
        user_id = row['user_id']
        song_id = row['song_id']
        if user_id not in user_songs:
            user_songs[user_id] = set()
        user_songs[user_id].add(song_id)
    return user_songs

def build_recommendation_model(train_data):
    """
    Build a recommendation model using collaborative filtering.

    Args:
    - train_data (DataFrame): DataFrame containing the training data with columns user_id, song_id, and rating.

    Returns:
    - model (surprise.prediction_algorithms.algo_base.AlgoBase): Trained recommendation model.
    """
    # Create a Surprise Dataset object
    reader = Reader(rating_scale=(0, 1))  # Since we only need binary rating (listened or not)
    data = Dataset.load_from_df(train_data[['user_id', 'song_id', 'rating']], reader)

    # Split the data into training and testing sets
    trainset, _ = train_test_split(data, test_size=0.0)  # Use full data for training
    
    # Build the recommendation model using KNNBasic algorithm (or any other algorithm of your choice)
    model = KNNBasic(sim_options={'user_based': False})  # Item-based collaborative filtering
    
    # Train the model
    model.fit(trainset)

    return model

# Preprocess the training data
user_songs = preprocess_train_data(train_data)

# Build the recommendation model
recommendation_model = build_recommendation_model(train_data)

print("Recommendation model built successfully!")
import pandas as pd
from sklearn.neighbors import NearestNeighbors
from scipy.sparse import csr_matrix

# Load the training data
train_data = pd.read_csv("train.csv")

# Preprocess the training data
user_songs = preprocess_train_data(train_data)

# Create a sparse user-item interaction matrix
user_item_matrix = train_data.pivot(index='user_id', columns='song_id', values='rating').fillna(0)
user_item_sparse = csr_matrix(user_item_matrix.values)

# Build the recommendation model using NearestNeighbors
model = NearestNeighbors(metric='cosine', algorithm='brute')
model.fit(user_item_sparse)

print("Recommendation model built successfully!")

''')