'''

Created on Mar 23, 2015

@author: mgulcin

'''

from collections import Counter

from collections import defaultdict

import gc

import operator

import os

import time

from Commons import Commons

class KItemsUnionRecommender(object):

def create_inverted_list(self, data):

index = defaultdict(list)

for i, tokens in enumerate(data):

for token in tokens:

index[token].append(i)

return index

'''

Difference from KNearestItems:

In KNI: Find items that are close to the target user only

In this (KIU): Find items that are close to the target user and his neighbors

'''

def find_rec_items(self, model, rec_neighbours, target_userid, unwantedList, k):

# I'm trying to do sth like do-while

postiveList = [target_userid]

postiveList.extend(rec_neighbours)

# remove unwanted items

commons = Commons()

items = commons.removeUnwantedItems( model = model, postiveList = postiveList,

topn = k, unwantedItemsList = unwantedList)

# find items to recommend

rec_items = items

# return

return rec_items

def getNeighbors(self, model, target_userid, unwantedList, N):

# I'm trying to do sth like do-while

rec_neighbours_tuples = model.most_similar(positive=[target_userid],topn=N)

rec_neighbours = [ seq[0] for seq in rec_neighbours_tuples ]

# if rec_items contain any locid(start with "loc") or "hometown", do something

s1, s2 = set(rec_neighbours), set(unwantedList)

shared = s1 & s2 # intersection, only the elements in both

while (len(rec_neighbours)-len(shared)) < N:

# increment output list size

temp_N = N + (len(shared)*3)

# get larger number of similar items

rec_neighbours_tuples = model.most_similar(positive=[target_userid],topn=temp_N)

rec_neighbours = [ seq[0] for seq in rec_neighbours_tuples ]

s1, s2 = set(rec_neighbours), set(unwantedList)

shared = s1 & s2 # intersection, only the elements in both

# remove items which start with "loc"

for loc_item in shared:

rec_neighbours.remove(loc_item)

# remove if too many items are collected

if len(rec_neighbours) > N:

rec_neighbours = rec_neighbours[0:N]

return rec_neighbours

def recommend(self, outFolderName, outFileName, model, users_list, loc_list, hometown_list, k=10, N=30):

rec_dict = dict()

commons = Commons()

temp_index = 0

user_collected_list = list()

for target_userid in users_list:

# get neighbours

unwantedList1 = loc_list + hometown_list

rec_neighbours = self.getNeighbors(model, target_userid, unwantedList1, N)

# get similar locations that neighbours has visited

unwantedList2 = users_list + hometown_list

rec_items = self.find_rec_items(model, rec_neighbours, target_userid, unwantedList2, k)

# add to dict

rec_dict[target_userid] = rec_items

user_collected_list.append(target_userid)

# print for every 10 target users

if (temp_index%100) == 0 and temp_index != 0:

commons.printRecs(rec_dict, user_collected_list, outFolderName, outFileName)

temp_index = 0

rec_dict.clear()

del user_collected_list[:]

else:

temp_index = temp_index + 1

#time.sleep(0.5) # to reduce cpu, or what should i do?

# print the rest

commons.printRecs(rec_dict, user_collected_list, outFolderName, outFileName)