import warnings
warnings.filterwarnings('ignore')

import numpy as np
import pandas as pd

import matplotlib.pyplot as plt
import seaborn as sns

from sklearn.metrics.pairwise import cosine_similarity

from sklearn.metrics import mean_squared_error

#Import the data set
df = pd.read_csv('/content/drive/MyDrive/ratings_Electronics.csv', header=None) #There are no headers in the data file

df.columns = ['user_id', 'prod_id', 'rating', 'timestamp'] #Adding column names

df = df.drop('timestamp', axis=1) #Dropping timestamp

df_copy = df.copy(deep=True) #Copying the data to another dataframe

counts = df['user_id'].value_counts()
df_final = df[df['user_id'].isin(counts[counts >= 50].index)]

# defining a function to get similar users
def similar_users(user_index, interactions_matrix):
    similarity = []
    for user in range(0, interactions_matrix.shape[0]): #  .shape[0] gives number of rows

        #finding cosine similarity between the user_id and each user
        sim = cosine_similarity([interactions_matrix.loc[user_index]], [interactions_matrix.loc[user]])

        #Appending the user and the corresponding similarity score with user_id as a tuple
        similarity.append((user,sim))

    similarity.sort(key=lambda x: x[1], reverse=True)
    most_similar_users = [tup[0] for tup in similarity] #Extract the user from each tuple in the sorted list
    similarity_score = [tup[1] for tup in similarity] ##Extracting the similarity score from each tuple in the sorted list

    #Remove the original user and its similarity score and keep only other similar users
    most_similar_users.remove(user_index)
    similarity_score.remove(similarity_score[0])

    return most_similar_users, similarity_score

# defining the recommendations function to get recommendations by using the similar users' preferences
def recommendations(user_index, num_of_products, interactions_matrix):

    #Saving similar users using the function similar_users defined above
    most_similar_users = similar_users(user_index, interactions_matrix)[0]

    #Finding product IDs with which the user_id has interacted
    prod_ids = set(list(interactions_matrix.columns[np.where(interactions_matrix.loc[user_index] > 0)]))
    recommendations = []

    observed_interactions = prod_ids.copy()
    for similar_user in most_similar_users:
        if len(recommendations) < num_of_products:

            #Finding 'n' products which have been rated by similar users but not by the user_id
            similar_user_prod_ids = set(list(interactions_matrix.columns[np.where(interactions_matrix.loc[similar_user] > 0)]))
            recommendations.extend(list(similar_user_prod_ids.difference(observed_interactions)))
            observed_interactions = observed_interactions.union(similar_user_prod_ids)
        else:
            break

    return recommendations[:num_of_products]