In [1]:
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
In [2]:
df = pd.read_table('./groceries.csv',header=None)
df1= df.iloc[:,0].str.get_dummies(sep=',')
products=df1.columns
df1.head()
Out[2]:
In [3]:
df1[(df1['baby food']>0) & (df1['whisky']>0)]
Out[3]:
In [4]:
from numpy import array
from numpy import diag
from numpy import dot
from numpy import zeros
from scipy.linalg import svd
import numpy as np
np.set_printoptions(formatter={'float': lambda x: "{0:0.3f}".format(x)})
In [5]:
A=np.array(df1)
#print(A)
In [6]:
A.shape
Out[6]:
In [7]:
U, L, VT = svd(A)
plt.plot(L*L)
Out[7]:
In [8]:
#lets verify that the sum of squares match
print("Sum of squares of singular values is",(L*L).sum())
print("Sum of squares of sales is",(A*A).sum())
In [24]:
-VT[0]
Out[24]:
In [25]:
#get product names with 5 largest values in firt row of VT
#note that all values are negative so I am asking for the bottom-5 values next
products[VT[0].argsort()[:5]]
Out[25]:
In [11]:
def generate_user(items):
user=np.zeros(A.shape[1])
for i in (items):
tf = products.get_loc(i)
user[tf]=100
return user
In [12]:
#CREATE A CUSTOMER John THAT LIKES tropical fruit
john=generate_user(['tropical fruit'])
print(john)
In [13]:
def what_do_you_like(c,products):
return products[c.argmax()]
print('John likes',what_do_you_like(john,products))
In [14]:
ATA=A.transpose().dot(A)
friends_of_john=ATA.dot(john)
print('Friends of John like',what_do_you_like(friends_of_john,products))
In [15]:
#print Friends-of-john vector
print(friends_of_john)
print('Product loc=',products.get_loc(what_do_you_like(friends_of_john,products)))
In [19]:
friends_of_friends_of_john=ATA.dot(friends_of_john)
print('Friends of friends of John like',what_do_you_like(friends_of_friends_of_john,products))