ML之RS：基于使用者的CF+LFM實作的推薦系統(基于相關度較高的使用者實作電影推薦)

輸出結果

實作代碼

#ML之RS：基于CF和LFM實作的推薦系統

import numpy as np

import pandas as pd

import matplotlib.pyplot as plt

import time

import warnings

warnings.filterwarnings('ignore')

np.random.seed(1)

plt.style.use('ggplot')

# data = pd.read_csv('ml-20m/ratings_smaller.csv', index_col=0)

# movies = pd.read_csv('ml-20m/movies_smaller.csv')

#1、導入資料集

data = pd.read_csv('ml-latest-small/ratings.csv')

movies = pd.read_csv('ml-latest-small/movies.csv')

movies = movies.set_index('movieId')[['title', 'genres']]

#2、觀察資料集

# How many users?

print (data.userId.nunique(), 'users')

# How many movies?

print (data.movieId.nunique(), 'movies')

# How possible ratings?

print (data.userId.nunique() * data.movieId.nunique(), 'possible ratings')

# How many do we have?

print (len(data), 'ratings')

print (100 * (float(len(data)) / (data.userId.nunique() * data.movieId.nunique())), '% of possible ratings')

# Number of ratings per users

fig = plt.figure(figsize=(10, 10))

ax = plt.hist(data.groupby('userId').apply(lambda x: len(x)).values, bins=50)

plt.xlabel("ratings")

plt.ylabel("users")

plt.title("Number of ratings per user")

plt.show()

# Number of ratings per movie

ax = plt.hist(data.groupby('movieId').apply(lambda x: len(x)).values, bins=50)

plt.ylabel("movies")

plt.title('Number of ratings per movie')

# Ratings distribution評分分布

ax = plt.hist(data.rating.values, bins=5)

plt.ylabel("numbers")

plt.title("Distribution of ratings")

# Average rating per user

ax = plt.hist(data.groupby('userId').rating.mean().values, bins=10)

plt.xlabel("Average rating")

plt.title("Average rating per user")

# Average rating per movie

ax = plt.hist(data.groupby('movieId').rating.mean().values, bins=10)

plt.title('Average rating per movie')

# Top Movies，genres電影類型

average_movie_rating = data.groupby('movieId').mean()

top_movies = average_movie_rating.sort_values('rating', ascending=False).head(10)

pd.concat([movies.loc[top_movies.index.values],

          average_movie_rating.loc[top_movies.index.values].rating], axis=1)

# Robust Top Movies - Lets weight the average rating by the square root of number of ratings讓平均評分進行權重數的平方根

top_movies = data.groupby('movieId').apply(lambda x:len(x)**0.5 * x.mean()).sort_values('rating', ascending=False).head(10)

pd.concat([movies.loc[top_movies.index.values],

controversial_movies = data.groupby('movieId').apply(lambda x:len(x)**0.25 * x.std()).sort_values('rating', ascending=False).head(10)

pd.concat([movies.loc[controversial_movies.index.values],

          average_movie_rating.loc[controversial_movies.index.values].rating], axis=1)