In [1]:
import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
%matplotlib inline
In [4]:
movie=pd.read_csv('movies.csv')
movie.head()
Out[4]:
In [7]:
df=pd.read_csv('ratings.csv')
df.head()
Out[7]:
Merging to create a new data frame¶
In [10]:
df=df.merge(movie,on='movieId')
In [11]:
df.head()
Out[11]:
In [13]:
df.drop('timestamp',axis=1,inplace=True)
In [14]:
df.head()
Out[14]:
In [15]:
df.shape
Out[15]:
In [31]:
rate_count=df.groupby('title')['rating'].count().sort_values(ascending=False)
rate_count
Out[31]:
Recommending user based on user ratings using correlation¶
In [24]:
movie_mat=pd.pivot_table(df,values='rating',columns='title',index='userId')
movie_mat.head(10)
Out[24]:
In [25]:
corr=movie_mat.corr()
Suggesting movie for Matrix, The (1999)¶
In [48]:
x=corr['Matrix, The (1999)'].sort_values(ascending=False)
x
Out[48]:
In [49]:
predict_df=pd.DataFrame({'correlation':x,'count':rate_count})
In [50]:
predict_df.head()
Out[50]:
In [51]:
predict_df.dropna(inplace=True)
In [55]:
print(predict_df[predict_df['count']>100].sort_values(by='correlation',ascending=False).head())
Recommending user using Collaborative Filtering¶
In [56]:
df.head()
Out[56]:
In [108]:
movie_mat=pd.pivot_table(df,values='rating',index='title',columns='userId').fillna(0)
movie_mat.head()
Out[108]:
In [128]:
movie_mat_new=movie_mat.reset_index()
In [130]:
movie_mat_new.head()
Out[130]:
In [60]:
from scipy.sparse import csr_matrix
In [61]:
vec_movie_mat=csr_matrix(movie_mat.values)
In [62]:
from sklearn.neighbors import NearestNeighbors
knn = NearestNeighbors(metric = 'cosine', algorithm = 'brute')
knn.fit(vec_movie_mat)
Out[62]:
Suggesting movie for Matrix, The (1999)¶
In [141]:
movie_mat_new[movie_mat_new['title']=='Matrix, The (1999)']
Out[141]:
In [142]:
x=movie_mat.iloc[5512,:].values.reshape(1,-1)
x
Out[142]:
In [145]:
distance,indices= knn.kneighbors(x, n_neighbors = 6)
print(distance,indices)
In [146]:
for i in range(len(distance.flatten())):
print(movie_mat.index[indices.flatten()][i],distance.flatten()[i])
In [ ]:
Recommending User based on content based filtering¶
In [147]:
df=pd.read_csv('tmdb_movies.csv')
In [148]:
df.head()
Out[148]:
In [149]:
df.shape
Out[149]:
In [151]:
df.info()
In [152]:
df.isnull().sum()
Out[152]:
In [154]:
df1=df[['original_title','overview']]
df1.head()
Out[154]:
In [259]:
df1.reset_index(inplace=True)
df1.head()
Out[259]:
In [157]:
from sklearn.feature_extraction.text import TfidfVectorizer
In [160]:
tf=TfidfVectorizer(stop_words='english')
In [162]:
tf_mat=tf.fit_transform(df1['overview'].fillna(''))
In [163]:
tf_mat.shape
Out[163]:
In [184]:
from sklearn.metrics.pairwise import cosine_similarity
In [186]:
cosine_sim = cosine_similarity(tf_mat)
In [284]:
df1[df1['original_title']=='The Dark Knight Rises']
Out[284]:
Suggesting movie for The Dark Knight Rises¶
In [287]:
similar=cosine_sim[3].tolist()
In [288]:
ind=[]
val=[]
for i in range(len(similar)):
ind.append(i)
val.append(similar[i])
In [289]:
dic={'index':ind,'value':val}
rec_df=pd.DataFrame(dic)
In [290]:
rec=rec_df.merge(df1,on='index')
rec.head()
Out[290]:
In [297]:
a=rec.sort_values(by='value',ascending=False).head(7)
In [298]:
print(a['original_title'])
