In [1]:
import sys
import pandas as pd
from helpers.plot_helper import PlotHelper

Load JSON dataset by pandas

In [2]:
ratings_sparseness_dataframe = pd.read_json('../data/ratings_sparseness.json')
ratings_sparseness_dict = ratings_sparseness_dataframe.to_dict()

Display dataset

In [3]:
ratings_sparseness_dataframe = ratings_sparseness_dataframe.transpose()
display(ratings_sparseness_dataframe)
m1 m2 m3 m4 m5 m6 m7 m8 m9 m10
saeed 4 2 5 - 2 3 4 5 - -
abbas 3 - 4 2 - - - - - -
alireza 3 - 4 5 2 3 4 5 5 5

distance metric implementation

In [4]:
def get_manhattan_distance(rating1,rating2):
    
    rating1 = dict(filter(lambda x: x[1] != '-', rating1.items()))
    rating2 = dict(filter(lambda x: x[1] != '-', rating2.items()))
    
    distance = 0
    for key in rating1:
        if key in rating2:
            distance += abs(rating1[key] - rating2[key])
    return distance

dt=ratings_sparseness_dict

print('saeed-alireza:%f'%get_manhattan_distance(dt['saeed'],dt['alireza']))
print('saeed-abbas:%f'%+get_manhattan_distance(dt['saeed'],dt['abbas']))

saeed-alireza:2.000000
saeed-abbas:2.000000
In [5]:
def get_euclidean_distance(rating1,rating2):
    
    
    rating1 = dict(filter(lambda x: x[1] != '-', rating1.items()))
    rating2 = dict(filter(lambda x: x[1] != '-', rating2.items()))
    
    distance = 0
    for key in rating1:
        if key in rating2:
            distance += pow(abs(rating1[key]-rating2[key]), 2)
    return pow(distance, 1/2)

dt=ratings_sparseness_dict

print('saeed-alireza:%f'%get_euclidean_distance(dt['saeed'],dt['alireza']))
print('saeed-abbas:%f'%+get_euclidean_distance(dt['saeed'],dt['abbas']))

saeed-alireza:1.414214
saeed-abbas:1.414214
In [6]:
def dot(A,B): 
    return (sum(a*b for a,b in zip(A,B)))

def cosine_similarity(rating1,rating2):
    rating1 = dict(filter(lambda x: x[1] != '-', rating1.items()))
    rating2 = dict(filter(lambda x: x[1] != '-', rating2.items()))
    a = rating1.values()
    b = rating2.values()
    return dot(a,b) / ( (dot(a,a) **.5) * (dot(b,b) ** .5) )

dt=ratings_sparseness_dict

print('saeed-alireza:%f'%cosine_similarity(dt['saeed'],dt['alireza']))
print('saeed-abbas:%f'%+cosine_similarity(dt['saeed'],dt['abbas']))

saeed-alireza:0.801784
saeed-abbas:0.559893