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]:
===== 2-03-21 Starting Over¶
In [2]:
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
sns.set_style('white')
In [15]:
# 1. Read in the data, change separator to tab
# 2. Create an array of new column names
# 3. Assign that array to our columns
df = pd.read_csv('u.data', sep="\t")
columns = ['user_id', 'item_id', 'rating', 'timestamp']
df.columns = columns
In [16]:
df.head()
Out[16]:
OH NO! How do we know what movies these are, just by the id!?¶
In [17]:
# 1. Read in the item_id to movie_title df
# 2. Merge
movietitles = pd.read_csv('Movie_Id_Titles')
df = pd.merge(df, movietitles, on="item_id")
Ok, how do we get JUST the ratings?¶
In [20]:
ratings = pd.DataFrame(df.groupby('title')['rating'].mean())
ratings.head()
Out[20]:
Ok, but, this is almost useless if only one person has rated the movie, right?¶
So, we should also get number of ratings, for posterity¶
In [22]:
ratings['number_of_ratings'] = df.groupby('title')['rating'].count()
In [24]:
ratings.head()
Out[24]:
PART TWO: Now we are on to recommending similar movies¶
In [29]:
# 1. Make a matrix (like an excel pivot table)
In [28]:
movie_matrix = df.pivot_table(index="user_id", columns="title", values="rating")
movie_matrix.head()
Out[28]:
In [30]:
# 2. Pick our movies
# 3. Get the ratings for only those movies
starwars_ratings = movie_matrix['Star Wars (1977)']
liarliar_ratings = movie_matrix['Liar Liar (1997)']
In [33]:
similar_to_starwars = movie_matrix.corrwith(starwars_ratings)
In [36]:
corr_with_starwars = pd.DataFrame(similar_to_starwars, columns=['correlation'])
corr_with_starwars
Out[36]:
In [40]:
sw_df = corr_with_starwars.join(ratings['number_of_ratings'])
In [42]:
sw_df = sw_df[sw_df['number_of_ratings'] >= 100 ]
In [44]:
sw_df = sw_df.sort_values('correlation', ascending=False)
In [45]:
sw_df
Out[45]:
Now, we're going to do the same thing with Liar Liar¶
- get all Liar Liar movies
- Get correlation
- Get number of ratings
In [46]:
df
Out[46]:
In [47]:
liar_liar = df[df['title'] == 'Liar Liar (1997)']
liar_liar
Out[47]:
In [48]:
df_matrix = df.pivot_table(index="user_id", columns="title", values="rating")
df_matrix
Out[48]:
In [49]:
df_m_sw = df_matrix['Star Wars (1977)']
In [53]:
df_m_sw_corr = df_matrix.corrwith(df_m_sw)
In [55]:
df_m_sw_corr = pd.DataFrame(df_m_sw_corr, columns=['correlation'])
df_m_sw_corr
Out[55]:
In [58]:
df_m_sw_corr_with_ratings = df_m_sw_corr.join(ratings)
In [59]:
df_m_sw_corr_with_ratings
Out[59]:
In [60]:
df_m_sw_corr_with_ratings = df_m_sw_corr_with_ratings[df_m_sw_corr_with_ratings['number_of_ratings'] >= 100]
In [61]:
df_m_sw_corr_with_ratings.sort_values('correlation',ascending=False)
Out[61]:
NOW with Liar Liar¶
In [81]:
# 1. Make a matrix
df_matrix = df.pivot_table(index="user_id", columns="title", values="rating")
# 2. Get JUST Liar Liar
df_matrix_ll = df_matrix['Liar Liar (1997)']
# 3. Compare everything in the matrix to Liar Liar column
df_matrix_ll_corr = df_matrix.corrwith(df_matrix_ll)
# 4. Turn that into a df so we can add things like 'ratings'
df_matrix_ll_corr_df = pd.DataFrame(df_matrix_ll_corr, columns=['correlation'])
# 5. Using a join (because our index -- title -- is the same for both dfs, add ratings)
df_matrix_ll_corr_df_with_ratings = df_matrix_ll_corr_df.join(ratings)
# 6. Remove anything with number of ratings < 100
df_matrix_ll_corr_df_with_ratings = df_matrix_ll_corr_df_with_ratings[df_matrix_ll_corr_df_with_ratings
['number_of_ratings'] >= 100]
# 7. Sort
df_matrix_ll_corr_df_with_ratings_sorted = df_matrix_ll_corr_df_with_ratings.sort_values('correlation',
ascending=False)
In [82]:
df_matrix_ll_corr_df_with_ratings_sorted
Out[82]:
In [78]:
# IN SHORT
dfm = df.pivot_table(index="user_id", columns="title", values="rating")
ll = dfm['Liar Liar (1997)']
dfm_ll = pd.DataFrame(dfm.corrwith(ll), columns=['correlation'])
dfm_ll = dfm_ll.join(ratings)
dfm_ll = dfm_ll[dfm_ll['number_of_ratings'] >= 100].sort_values('correlation', ascending=False)
dfm_ll
Out[78]:
In [ ]: