Page 149 - ITU KALEIDOSCOPE, ATLANTA 2019

ICT for Health: Networks, standards and innovation

Table 1 – The fuzzy number of scoring set Table 2 – Unknown data for global optimization

Intuitionistic Fuzzy Numbers SymbolMeaningSymbol Meaning
Level
[µ QR ,ν QR − ρ QR ] Unknown M Quantity V Performance Values
Excellent [0.9, 0.1-ρ QR ] Data W Weight L Installation Location
Good [0.7, 0.3-ρ QR ]
Medium [0.5, 0.5-ρ QR ]
Poor [0.3, 0.7-ρ QR ] ( A Õ Õ ! )
B

Very Poor [0.1, 0.9-ρ QR ] Y 1 = Min p ij ∗ x ij ∗ m i
i=1 j=1
B
( ! )
A Õ Õ
q QR = µ QR − ν QR × ρ QR (9) Y 2 = Max v ij ∗ x ij ∗ w i
i=1 j=1
In order to compare performance of smart services, the QoS B !
values need to be normalized. In this paper, Equation (10) in  A Õ Õ (14)



[21] is adopted to normalize positive QoS attributes of smart  j=1 p ij ∗ x ij ∗ m i ≤ Max_ Cost,


 i=1

services, and Equation (11) is used for negative attributes. 


 B Õ
 
x ij = 1,i = 1,2,..., A
s.t.
q i j −q min
 
 j max min  j=1

  min i f q − q , 0
q ij = q max −q j j (10)  (

j
j

 1 i f q max − q min = 0  1, sensing device x ij is selected


j j 

 x ij =

 0, sensing device x ij is not selected
( q max −q i j 
j min i f q max − q min , 0
q ij = q max −q j j j (11) 5.2 Unknown data acquisition
j
1 i f q max − q min = 0
j j
Then, the cosine similarity is used to calculate the As opposed to web services and cloud services, the cost
non-functional similarity between care demand c i and smart of smart services depends on the price of sensing devices.
service s j , as shown in Equation (12), where Q represents However, two different services may require the same sensing
0
the normalized QR and Q is the normalized QoS. device. For example, the Activity Monitoring service requires
the acceleration data and the Falling Detection service
requires the same data. Therefore, we need to reduce
repeatable sensing devices of selected services.
5 Í
q 0 k × q k
Q Q 0 k=1
q_sim(c i , s j ) = = s s (12) Definition 3 As a sensing device in smart service work in a
0
||Q|| × ||Q || specific position, we formalize every required sensing device
5 Í
q 02 k × 5 Í q k 2 with its type and position, as d i = {type, position}.
k=1 k=1
In Definition 3, type represents the type of the sensing device
Finally, the mapping similarity between atomic care demand and position is the installation position. Therefore, we count
c i and smart service s j is calculated by Equation (13), based the quantity of every type of sensing device with two rules:
on functional similarity and non-functional similarity, where
γ 1 and γ 2 are the weight coefficients with γ 1 + γ 2 = 1. 1. If d i and d j from different services in S desire have
the same d_type and d_position, we suppose that one
sensing device is enough for both smart services. Thus,
the quantity of this type of sensing devices is unchanged,
m_sim(c i , s j ) = γ 1 × s_sim(c i , s j )+γ 2 ×q_sim(c i , s j ) (13)
while the weight is increased by one.
5. GLOBAL OPTIMIZATION ALGORITHM 2. If d i and d j from different services in S desire have the
same d_type but different d_positions, only one sensing
5.1 Transformation of sensing devices selection
device is not enough. Hence, the quantity and weight of
In order to improve service performance and reduce the cost this type of sensing devices are both increased by one,
of sensing devices, we first need to count the requirement of and two installation locations are added into the L.
sensing devices for geriatric care. Suppose that there are A In this paper, we adopt Formula (15) to evaluate the
types of sensing devices and B commodities for each type of performance of sensing devices, where ω is the compensation
sensing devices. The problem of selecting sensing devices coefficient and ξ is the general error. Additionally, y is the
can be converted into a multi-objective knapsack problem, service life of sensing devices, r is the measuring range and
if we can calculate the unknown data in Table 2. Then, we r is the average measuring range. η (η ∈ (0,1)) indicates the
establish a global optimization selection model to maximize effect of sensing devices in smart services.
the total performance and to minimize the total cost of sensing
devices, as shown in Equation (14). Therefore, this problem ωy r r
is a NP-complete problem like the knapsack problem. R = η (15)
ξ r
– 129 –