ass.csv

recommend.py

import csv
import numpy as np
import math

def col(i,lis):
    return [r[i] for r in lis][:-1]

def predict(l,n,d,idf=1):
    profile1, profile2, docscore1,docscore2 = [],[],[],[]
    for i in range(1,d-3):
        c = col(i,l)
        profile1.append(np.dot(c,col(d-2,l)))
        profile2.append(np.dot(c,col(d-1,l)))
    print ("profiles")
    print (profile1)
    print (profile2)
    if idf is 1:
        docscore1 = [np.dot(profile1,r[1:-3]) for r in l[:-1]]
        docscore2 = [np.dot(profile2,r[1:-3]) for r in l[:-1]]
    else:
        print ("debug",len(idf),len(profile1))
        docscore1 = [sum([x*y*z for x,y,z in zip(profile1,r[1:-3],idf)]) for r in l[:-1]]
        docscore2 = [sum([x*y*z for x,y,z in zip(profile2,r[1:-3],idf)]) for r in l[:-1]]

    print ("doc scores")
    print (docscore1)
    print (docscore2)
    print ("user1 will like document {d} (score = {s}) best".format(d=docscore1.index(max(docscore1))+1,s=max(docscore1)))
    print ("user1 will like document {d} (score = {s}) 2nd best".format(d=docscore1.index(sorted(docscore1)[-2])+1,s=sorted(docscore1)[-2]))
    print ("user1's prediction for doc9 is",docscore1[8])
    print ("user2 will like document {d} (score = {s}) best".format(d=docscore2.index(max(docscore2))+1,s=max(docscore2)))
    print ("No. of docs user2 will dislike is ",sum(1 for x in docscore2 if x < 0))


with open('ml-latest-small/ass.csv') as f:
   data = csv.reader(f,delimiter=',')
   next(data)
   l = list(data)
   n = len(l)
   d = len(l[0])

   for i in range(n):
       for j in range(1,d):
           l[i][j] = int(l[i][j]) if l[i][j]!='' else 0

   predict(l,n,d)

   for r in l[:-1]:
       r[1:-3] = [x/math.sqrt(r[-3]) for x in r[1:-3]]

   predict(l,n,d)
   idf = [1/x for x in l[-1][1:-3]]

   predict(l,n,d,idf)