import numpy as np

A = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]])
print(A)
# Singular-value decomposition
U, s, VT = np.linalg.svd(A)
# create n x n Sigma matrix
Sigma = np.diag(s)
# reconstruct matrix
PT = Sigma.dot(VT)
#B = U.dot(Sigma.dot(VT))
print(PT)

# Singular-value decomposition 
A = np.array([[1, 2], [3, 4], [5, 6]])
U, s, VT = np.linalg.svd(A)
# create n x n Sigma matrix
Sigma = np.zeros((A.shape[0], A.shape[1]))
# populate Sigma with n x n diagonal matrix
Sigma[:A.shape[1], :A.shape[1]] = np.diag(s)
# reconstruct matrix
PT = Sigma.dot(VT)
B = U.dot(PT)

print('A = \n', A, '\n')
print('U = \n', U, '\n')
print('Sigma = \n', Sigma, '\n')
print('VT = \n', VT, '\n')
print('PT = \n', PT, '\n')
print('B = \n', B, '\n') 

# Singular-value decomposition
A = np.array([[1, 2, 3], 
              [4, 5, 6]])
U,S,VT = np.linalg.svd(A)
# create n x n Sigma matrix
Sigma = np.zeros((A.shape[1], A.shape[1]))
# populate Sigma with n x n diagonal matrix
if A.shape[1] > S.shape[0]:
    S = np.append(S, 0)
Sigma[:A.shape[1], :A.shape[1]] = np.diag(S)

PT= Sigma.dot(VT)
PT = PT[0:A.shape[0]]
B = U.dot(PT)
print('A = \n', A, '\n')
print('U = \n', U, '\n')
print('Sigma = \n', Sigma, '\n')
print('VT = \n', VT, '\n')
print('PT = \n', PT, '\n')
print('B = \n', B, '\n')

# https://beckernick.github.io/matrix-factorization-recommender/
import pandas as pd
import numpy as np

ratings_list = [i.strip().split("::") for i in open('/Users/datalab/bigdata/cjc/ml-1m/ratings.dat', 'r').readlines()]
users_list = [i.strip().split("::") for i in open('/Users/datalab/bigdata/cjc/ml-1m/users.dat', 'r').readlines()]
movies_list = [i.strip().split("::") for i in open('/Users/datalab/bigdata/cjc/ml-1m/movies.dat', 'r', encoding = 'iso-8859-15').readlines()]

ratings_df = pd.DataFrame(ratings_list, columns = ['UserID', 'MovieID', 'Rating', 'Timestamp'], dtype = int)
movies_df = pd.DataFrame(movies_list, columns = ['MovieID', 'Title', 'Genres'])

movies_df['MovieID'] = movies_df['MovieID'].astype('int64')
ratings_df['UserID'] = ratings_df['UserID'].astype('int64')
ratings_df['MovieID'] = ratings_df['MovieID'].astype('int64')

movies_df.head()

ratings_df.head()

# 注意：使用0填充缺失值
R_df = ratings_df.pivot(index = 'UserID', columns ='MovieID', values = 'Rating').fillna(0)
R_df.head()

R = R_df.to_numpy(dtype=np.int16)
user_ratings_mean = np.mean(R, axis = 1)
R_demeaned = R - user_ratings_mean.reshape(-1, 1)

from scipy.sparse.linalg import svds
U, sigma, Vt = svds(R_demeaned, k = 50)

sigma = np.diag(sigma)

all_user_predicted_ratings = U.dot( sigma.dot(Vt)) + user_ratings_mean.reshape(-1, 1)
preds_df = pd.DataFrame(all_user_predicted_ratings, columns = R_df.columns)

preds_df
# each row is a user
# each column is a movie

def recommend_movies(preds_df, user_row_number, movies_df, ratings_df, num_recommendations=5):
    # Get and sort the user's predictions
    sorted_user_predictions = preds_df.iloc[user_row_number].sort_values(ascending=False)
    
    # Get the user's data and merge in the movie information.
    userID = user_row_number + 1
    user_data = ratings_df[ratings_df.UserID == userID]
    user_full = (user_data.merge(movies_df, how = 'left', left_on = 'MovieID', right_on = 'MovieID').
                     sort_values(['Rating'], ascending=False)
                 )

    print('UserID {0} has already rated {1} movies.'.format(userID, user_full.shape[0]))
    print('Recommending the highest {0} predicted ratings movies not already rated.'.format(num_recommendations))
    
    # Recommend the highest predicted rating movies that the user hasn't seen yet.
    potential_movie_df= movies_df[~movies_df['MovieID'].isin(user_full['MovieID'])]
    predicted_movie_df = pd.DataFrame(sorted_user_predictions).reset_index()
    predicted_movie_df['MovieID'] = predicted_movie_df['MovieID'].astype('int64')
    recommendations = (
        potential_movie_df.merge(predicted_movie_df, how = 'left', on = 'MovieID').
         rename(columns = {user_row_number: 'Predictions'}).
         sort_values('Predictions', ascending = False).
                       iloc[:num_recommendations, :-1]
                      )

    return user_full, recommendations 

already_rated, predictions = recommend_movies(preds_df, 0, movies_df, ratings_df, 10)

already_rated[:3]

predictions