ratings.py

import numpy as np

class ratings(object):
 """ratings"""

 def __init__(self, inputdict):
  self._data = inputdict
  self.get_person_ids()
  self.get_paper_ids()
  self.get_ratings_array()

 def get_person_ids(self):
  self.persons = self._data.keys()
  self.persons.sort()

 def get_paper_ids(self):
  self.papers = []
  for person in self.persons:
   for paper in self._data[person].keys():
    if paper not in self.papers:
     self.papers.append(paper)
  self.papers.sort()

 def get_ratings_array(self):
  self.myratings = np.zeros((len(self.persons),len(self.papers)), dtype = float)
  for person in self.persons:
   for paper in self._data[person].keys():
    self.myratings[self.persons.index(person),self.papers.index(paper)] = self._data[person][paper]

 def compare_ratings(self, person1, person2):
 # only check non-zero ratings
  indices = np.where(self.myratings[self.persons.index(person1)] != 0) or np.where(self.myratings[self.persons.index(person2)] != 0)
  if len(indices) == 0: return len(self.papers)*25+1 #5 is max difference;no basis to compare
  cmp1 = self.myratings[self.persons.index(person1)][indices]
  cmp2 = self.myratings[self.persons.index(person2)][indices]
  return np.linalg.norm(cmp1 - cmp2)

 def recommendations(self):
  allrecommendations = {}
  for person in self.persons:
   minimum = len(self.papers)*25+1
   notread = np.array([])
   for otherperson in self.persons:
    if otherperson == person: continue;
    tmp = self.compare_ratings(person, otherperson)
    if tmp < minimum:
     closest = otherperson
     minimum = tmp
     notread = np.where(np.array(self.myratings[self.persons.index(otherperson)] == 0))[0]
   if len(notread) == 0:
    allrecommendations[person] = "No recommendation"
   else:
    maxindex = np.where(self.myratings[self.persons.index(otherperson)][notread] == self.myratings[self.persons.index(otherperson)][notread].max())
    allrecommendations[person] = self.papers[int(maxindex[0])]
  return allrecommendations

# Alternative distances:
# pearson
# tanimoto scipy.spatial.distance.rogerstanimoto