Recommender¶
In [2]:
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
sns.set_style('white')
In [3]:
df = pd.read_csv('u.data', sep="\t")
columns = ['userid', 'item_id', 'rating', 'timestamp']
df.columns = columns
In [4]:
movietitles = pd.read_csv('Movie_Id_Titles')
In [5]:
df = pd.merge(df, movietitles, on='item_id')
Make a "ratings" df¶
In [6]:
df.groupby('title')['rating'].mean()
Out[6]:
In [7]:
ratings = pd.DataFrame(df.groupby('title')['rating'].mean())
In [8]:
ratings['number_of_ratings'] = df.groupby('title')['rating'].count()
In [10]:
sns.jointplot(x="rating", y="number_of_ratings", data=ratings, alpha=0.5)
Out[10]:
Recommending Similar Movies¶
In [13]:
movie_matrix = df.pivot_table(index="userid", columns="title", values="rating")
movie_matrix
Out[13]:
Sort the ratings¶
In [14]:
ratings.sort_values('number_of_ratings', ascending=False).head(10)
Out[14]:
In [15]:
starwars_user_ratings = movie_matrix['Star Wars (1977)']
In [16]:
liarliar_user_ratings = movie_matrix['Liar Liar (1997)']
corrwith()¶
In [17]:
movie_matrix.corrwith(starwars_user_ratings)
Out[17]:
In [20]:
similar_to_starwars = movie_matrix.corrwith(starwars_user_ratings)
In [21]:
corr_starwars = pd.DataFrame(similar_to_starwars, columns=['Correlation'])
corr_starwars.dropna(inplace=True)
In [22]:
corr_starwars.head()
Out[22]:
Remove movies with less than 100 reviews¶
In [23]:
corr_starwars = corr_starwars.join(ratings['number_of_ratings'])
In [24]:
corr_starwars
Out[24]:
In [ ]: