View
221
Download
0
Category
Preview:
Citation preview
A Restaurant Recommendation System using Pearson
Correlation for Similarity Measurement
Present by : Arif Akbarul Huda, S.Si, M.Eng
background
2011 2012 2013 20140
5000
10000
15000
20000
25000
30000
35000
The growth of restaurant in Indonesia
http://www.statista.com/statistics/254456/number-of-internet-users-in-indonesia/
INFORMATION OVERLOAD
For some people, the overload of information become difficult specially how to manage it and finding which we need.
(A. Zanda, 2012)
SOLUTION
Recommender Systems
RELATED WORK
Nugroho [1] telah berhasil membuat sistem rekomendasi restoran dengan mempertimbangkan kebutuhan kalori penggunanya. 20014
Arief [8] membangun sebuah sistem rekomendasi wisata di
wilayah Yogyakarta dengan pendekatan collaborative-filtering
dan penyaringan informasi berdasarkan lokasi. 2012
Chu [5] memperkenalkan sistem rekomendasi restoran yang dapat mengenali perilaku diet,
pola makan dan kesukaan makanan bersayur pada penggunanya. 2013
Daraghmi [14] memberikan kontrobusi dalam pembuatan sistem rekomendasi restoran dengan
mempertimbangkan agama, budaya, alergi, riwayat kesehatan, dan aktifitas diet
penggunanya 2013
Liu [13] memerkenalkan teknik rekomendasi restoran dengan cara memperhatikan opini dan rating yang
diberikan penggunanya. 2013
2011 2012 2013 2014 2015
In fact, the importance of taste allow person to select a food in accord
with desires (Arthur Guyton)
The Fact of Taste
✔ Taste => chemical reaction
✔ 5 basic of taste => Sweet, Sour, Salty,
Umami, Bitter
Arthur Guython (Textbook of Medical Physiology, page 665)
Proposed Algorithm
attribute
BitterSalty
SavorySour
SweetSauceSpicyMeat
vegetable
BitterSalty
SavorySour
SweetSauceSpicyMeat
vegetable
User preferenceFoods taste character
attributeitem
Soto Ayam Kampung
User preference
bitter 0.00 0.00
sweet 0.70 0.63
savory 0.60 0.60
salty 0.20 0.23
sour 0.00 0.07
spicy 0.00 0.43
sauce 1.00 0.67
meat 0.80 0.83
vegetable 0.70 0.57
attribute
foodsUser
preferenceMie Ayam Super Jumbo
Soto Ayam Kampung
Soto Campur
Rica-rica Mentok
Tengkleng Kambing
bitter 0.00 0.00 0.00 0.00 0.00 0.00
sweet 0.50 0.70 0.70 0.40 0.40 0.63
savory 0.40 0.60 0.50 0.40 0.40 0.60
salty 0.20 0.20 0.20 0.10 0.10 0.23
sour 0.00 0.00 0.20 0.00 0.10 0.07
spicy 0.50 0.00 0.40 0.70 0.60 0.43
sauce 0.80 1.00 1.00 0.70 0.60 0.67
meat 0.80 0.80 0.60 1.00 1.00 0.83
vegetable 0.30 0.70 0.70 0.00 0.00 0.57
Which food should be recommend?
Similarity measurement
● Eudiclane distance● Cosine similarity● Pearson Corellation
“Pearson correlation tends to give better result in situation where data not well normalized.” (T. Segaran,2007)
Pearson Correlation
Similarity measurement
atributitem
Soto Ayam Kampung
preferensi pengguna
bitter 0.00 0.00
sweet 0.70 0.63
savory 0.60 0.60
salty 0.20 0.23
sour 0.00 0.07
spicy 0.00 0.43
sauce 1.00 0.67
meat 0.80 0.83
vegetable 0.70 0.57
atributitem
x y
bitter 0.00 0.00
sweet 0.70 0.63
savory 0.60 0.60
salty 0.20 0.23
sour 0.00 0.07
spicy 0.00 0.43
sauce 1.00 0.67
meat 0.80 0.83
vegetable 0.70 0.57
Similarity measurement
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.20
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
bitter
sweetsavory
salty
sour
spicy
sauce
meat
vegetable
Soto Ayam Kampung
pre
fere
nsi
pe
ng
gu
na
r(soto ayam, preferensi pengguna)=0.789
√0.658∗√1.242=0.8727
Finding a similarity...
Range value note message
0.80-1.00 Very high correlation recommended
0.60-0.79 korelasi tinggi Maybe you liked
0.40-0.59 Low correlation Try it
0.20-0.39 Very low correlation -
0.00-0.19 No correlation -
<< -1.00 berkebalikan -
The range of correlation
Adityo Nugroho
● 2 year chef at Hotel Royal Ambarukmo● 4 year chef at Hotel 101 Yogyakarta● 2 year as menu consultant at The Real
Steak House Batam● for 3 years till now has own catering
our expert….
Conclusion
● Pearson Correlation formula can be used for finding a recommended foods appropriate to user preference
● A food need to be extracted into nine attributes to identify its characteristic
Thank you
Recommended