Expert Systems with Applications 41 (2014) 81898202

Contents lists available at ScienceDirect

Expert Systems with Applications

journal homepage: www.elsevier.com/locate/eswa

A genetic algorithm approach for multi-product multi-period

continuous review inventory models
Ilkay Saracoglu a,, Seyda Topaloglu b, Timur Keskinturk c
a

Graduate School of Natural and Applied Sciences, Dokuz Eylul University, Tinaztepe Campus, 35160 Izmir, Turkey
Department of Industrial Engineering, Dokuz Eylul University, Tinaztepe Campus, 35160 Izmir, Turkey
c
Department of Quantitative Methods, School of Business, Istanbul University, Avcilar, 34180 Istanbul, Turkey
b

a r t i c l e

i n f o

Article history:
Available online 14 July 2014
Keywords:
Multi-product multi-period policy
Continuous review policy
Inventory management
Variable demand

a b s t r a c t
This paper formulates an approach for multi-product multi-period (Q , r) inventory models that calculates
the optimal order quantity and optimal reorder point under the constraints of shelf life, budget, storage
capacity, and extra number of products promotions according to the ordered quantity. Detailed
literature reviews conducted in both elds have uncovered no other study proposing such a multiproduct (Q , r) policy that also has a multi-period aspect and which takes all the aforementioned constraints into consideration. A real case study of a pharmaceutical distributor in Turkey dealing with large
quantities of perishable products, for whom the demand structure varies from product to product and
shows deterministic and variable characteristics, is presented and an easily-applicable (Q , r) model for
distributors operating in this manner proposed. First, the problem is modeled as an integer linear
programming (ILP) model. Next, a genetic algorithm (GA) solution approach with an embedded local
search is proposed to solve larger scale problems. The results indicate that the proposed approach yields
high-quality solutions within reasonable computation times.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
Customer satisfaction has recently become one of the most
important issues for companies as a consequence of globalization,
increased competition, and shrinking prot margins. The most fundamental condition associated with customer satisfaction is the
availability of products upon request. In order for companies to
satisfy this condition, excellent inventory management is essential.
Much research has been conducted on inventory management,
and results have shown that the type of inventory management
needed varies according to the nature of the companies. To construct an inventory management policy, it is therefore necessary
to conduct an in-depth analysis of the nature of the target company. The inventory manager must determine (1) when an order
should be placed, and (2) how much should be ordered. Further,
inventory, production, marketing, and nance managers should
operate in concert in order to reach an agreement on how to
reduce production costs and inventory investments, and increase
customer responsiveness.
Good inventory management is closely linked to the inventory
policy implemented, and is based upon a profound analysis of
Corresponding author. Tel.: +90 5324520069; fax: +90 2123665802.
E-mail address: ilkays@sbbdanismanlik.com (I. Saracoglu).
http://dx.doi.org/10.1016/j.eswa.2014.07.003
0957-4174/ 2014 Elsevier Ltd. All rights reserved.

the system. Harris (1913) rst estimated the economic order quantity to minimize the total cost of keeping stocks and placing orders.
A model was developed for a single product without any
constraints on the assumption that demand is known and lead
time is zero.
However, in many real inventory systems, there is usually
uncertainty with regard to demand and lead time. Continuous
review inventory policy (Q , r) is one of the most common policies
for cases in which the demand is uncertain and a supply lead time
exists. In this policy, an order quantity Q is placed when the inventory level is at or below reorder level r. The (Q , r) policies applied
on single items with xed lead time have been studied by
researchers such as Hadley and Whitin (1963), Federgruen and
Zheng (1992), Zheng (1992), Axster (1993), Kim and Benton
(1995), Axster (2007), Lau and Lau (2008), Mattsson (2010),
Hajiaghaei-Keshteli, Sajadifar, and Haji (2011), and Ang, Song,
Wang, and Zhang (2013).
In actual inventory systems, however, multi-item storage is an
issue typically encountered. For instance, wholesalers, large
distribution chains, and department stores manage an inventory
system comprising a wide range of products. Garman (1976),
Gardner (1983), Hopp, Spearman, and Zhang (1997) formulated
multi-product (Q , r) inventory policy with some constraints such
as inventory investment, service level and order frequency. Jeddi,

8190

I. Saracoglu et al. / Expert Systems with Applications 41 (2014) 81898202

Shultes, and Haji (2004) analyzed multi-product stochastic inventory system with backorders, shortages, and budget constraints.
Betts and Johnston (2005) constructed a multi-item (Q , r) model
with the objective of determining an optimal replenishment policy
by reducing risks and increasing prot.
As in most of the inventory management problems, unfortunately, the model is naturally complex for the multi-product case
and it is intractable to control optimal policies analytically due to
the size of the solution domain. Therefore, it is observed that iterative procedures such as heuristics and metaheuristics are used in
order to reach a solution in multi-product inventory systems.
Pasandideh, Niaki, and Tokhmehchi (2011) developed a multi-item
(Q , r) inventory policy for a two-echelon inventory system by using
genetic algorithm (GA). Pasandideh, Niaki, and Nia (2011)
proposed a multi-product EOQ model which does not only consider
the storage capacity, but also assumes the number of orders limited, and they used GA to solve it. Mandal, Maity, Maity, Mondal,
and Maiti (2011) formulated multi-item multi-period production
problems with fuzzy-random parameters and space constraints
and solved using GA. Yang, Chan, and Kumar (2012) developed a
GA approach for solving the multi-retailer, multi-product and
multi-period inventory system with backlogging and transportation capacity. Mousavi, Hajipour, Niaki, and Alikar (2013)
suggested a heuristic using GA and simulated annealing for the
multi-item multi-period inventory control problem with
discounts, time value of money, and ination.
This paper analyzes a pharmaceutical distributor that buys
medicines from pharmaceutical manufacturers and resells them
to pharmacies. The objective of the distributor is to meet the
demands of the pharmacies on time. Because a large number of
distributors is available in the sector, competition is intense.
Consequently, customer satisfaction is viewed with the utmost
priority. However, because distributors have to operate within
the connes of budget constraints, they need to make optimum
use of their budgets. In addition to the budget, shelf life and storage
area for drugs that require cold storage are additional
constraints that have to be satised. Further, pharmaceutical manufacturers may desire to deliver an additional amount of drugs for
free to induce increases in the number of medicines sold in the
pharmacies, instead of offering discounts on the basis of quantity
ordered. A further item to bear in mind is the fact that pharmaceutical manufacturers may provide convenient payment options to
the distributor who, in turn, may pass on these payment options
to the pharmacies. Regarding the pharmaceutical sector, Bijvank
and Vis (2012) provided an overview on the literature for hospital
inventory systems. Lapierre and Ruiz (2007) modeled the
multi-product, multi-period logistics system as a mixed-integer
program under the limitation of storage and manpower capacities,
and developed a tabu search to solve the problem owing to the
large scale of the problem.
This paper constructs a multi-item multi-period (Q , r) inventory
policy considering items with variable demand and determines the
optimal order quantity (Q) and reorder point (r) for each item to
maximize the prot for the pharmaceutical distributor in question,
considering all these constraints.
When we examine the companies in the pharmaceutical sector,
it can be seen that they are using a particular database and the
functionality of their systems is no more than keeping track of past
sales and current stock information. Yet they are not capable of
performing an optimization like the one we are suggesting in this
study. To give an example, among one of the most developed
Enterprise Resource Planning (ERP) programs, SAP uses Materials
Requirements Planning (MRP) and Distribution Requirements
Planning (DRP). MRP makes calculation for the dependent demand
by taking the input value (such as days of supply, lead time, forecast, Master Production Scheduling (MPS), safety stock) introduced

to the system as a base and provides purchase of needed amount

and determines order time by considering lead time, whereas
DRP system performs the distribution planning for the customers.
While advanced planning and scheduling (APS) models are embedded to SAP to perform optimization using mathematical programming techniques, these models work under the assumption that all
parameter values are constant and known absolutely (Kallrath &
Maindl, 2006, Louly & Dolgui, 2013), they do not address the
specic inventory management problem in this study.
The remainder of this paper is organized as follows: Literature
review is presented in Section 2. Problem description and assumptions, and our proposed mathematical model are given in Section 3.
A GA based solution approach is developed to solve the problem in
Section 4. An illustrative computational case study is given in Section 5. In order to demonstrate how the proposed methodology is
applied, a real case study and sensitivity analysis are carried out in
Section 6. Finally, conclusions and recommendations for future
research are provided in Section 7.

2. Review of current literature

In this paper, a multi-item perishable inventory management
system is proposed. Many products may lose their market value
over time as they become outdated, called perishability. Perishability is divided into two categories: xed and random lifetime. Fixed
lifetime is used for items that have a specic physical lifetime, such
as photographic lm, drugs, vitamins, and ready packed food and
milk. Random lifetime is used for items whose spoilage time
cannot be estimated, such as fresh vegetables, fruits, and meats.
The xed lifetime and exponential decay inventory study performed by Van Zyl (1964) is among the rst perishable items study
conducted. Nahmias (1982), Nahmias (2011, 1991) and Goyal and
Giri (2001) reviewed the current literature on ordering policies for
perishable items.
Continuous review policy (Q , r) has been used for nonperishable
items in a number of studies. The rst such study was conducted
by Berk and Gurler (2008). The results of our literature survey performed to date show the (Q , r) inventory model being used as single period in multi-product problems and for nonperishable
products, as illustrated in Table 1.
Zhao, Fan, Liu, and Xie (2007) tackled optimization with
storage-space constraint for the single-item (Q , r) policy, and proposed an efcient algorithm with polynomial time computational
complexity that obtains the optimal solutions. Then, they extended
their proposed method to enhanced multi-item (Q , r) policies.
Wang and Hu (2010) studied a continuous review (Q , r) model to
nd the optimal lot size and reorder point for a multi-item inventory model with interactions among items under budgetary constraints. They formulated the problem as a nonlinear one and
constructed a simple heuristic approximating procedure to solve
it. Kundu and Chakrabarti (2012) analyzed uncertain lead time
and demand condition for multi-item production. They created a
mathematical model that can determine the safety factor and
order quantity and proposed a continuous review model for a
multi-product system with both backorders and lost sales under
budget constraints. Zhao, Qiu, Xie, and He (2012) investigated single- and multi-item (Q , r) inventory systems with a stochastic
demand, limited resources, constant lead time, and backorders.
We discovered from our literature review that the (Q , r) inventory policy has not yet been carried out as multi-item multi-period.
The multi-period studies reviewed primarily dealt with lot-sizing
problems in production. Veinott (1965) considered a dynamic nonstationary multi-product inventory model in which the system was
reviewed in equal-length time periods, unsatised demand was
backlogged, and deterioration of stock did not occur in storage.

8191

I. Saracoglu et al. / Expert Systems with Applications 41 (2014) 81898202

Table 1
Published literature dealing with the multi-item (Q, r) inventory policy.
Author, year
Multi item
Single period
Nonperishable item
Garman (1976)
Gardner (1983)
Hopp et al. (1997)
Jeddi et al. (2004)
Betts and Johnston
(2005)
Zhao et al. (2007)
Wang and Hu (2010)
Pasandideh et al.
(2011)
Kundu and
Chakrabarti
(2012)
Zhao et al. (2012)
Multi period
Perishable item
Developed model in
this paper

Budget

Service
level

Storage

Inventory
investment

Order
frequency

Backorder

Resource
constraint

Extra
quantity

Lead
time

Demand

Method

F
F
F
F
F

St
D
St
St
D/St

L/H
L/H
L/H
L/H
A

F
F
F

St
St
D

St/H
H
GA

St

St

I/LS

St, D, Se

GA/LS

Lifetime

St: Stochastic; Se: Seasonal; D: Deterministic; F: Fixed; A: Analytical; H: Heuristics; I: Iterative Procedure; L: Lagrangian; GA: Genetic Algorithm; LS: Local Search; R: Random.

The studies dealing with multi-item multi-period models are

summarized in Table 2. Zhang and Eksioglu (2009) performed the
sole multi-period study for perishable items. In their study, they
presented a mixed-integer programming (MIP) model with
multi-mode replenishment options and perishable inventory that
can be stored for a limited number of periods and considered a single item.
The primary objective of this paper is to develop a novel integer
linear programming (ILP) model for the multi-product multi-period perishable products (Q , r) inventory problem with deterministic and variable demand and under budget, storage space, shelf life,

and extra number of products promotions constraints. Because

of the large number of decision variables and constraints in the
proposed model, it is difcult to solve this model via exact methods, even for small problems. Consequently, to nd a multi-item
multi-period (Q , r) inventory policy that tries to maximize the total
prot, we propose a GA solution approach.
3. Problem statement and formulation
We focus on a pharmaceutical distributor that collects products
from pharmaceutical manufactures and distributes them to

Table 2
Published literature dealing with multi-item multi-period inventory models.
Author, year

Lifetime

Demand

Excess
demand

Method

Policy

Lead
time

St

Backlogging

St

Gen and Syarif

(2005)
Kannan et al. (2010)

HGA

Mandal et al. (2011)

No
shortages
No
shortages
Shortages

Pasandideh et al.
(2013)
Mousavi et al.
(2013)
Single item + multi
period
Perishable
Zhang and Eksioglu
(2009)
Multi-item + multiperiod
Perishable
Developed model in
this paper

No
shortages
No
shortages

GA

Optimal
ordering
Optimal
ordering
Supply
chain
Production
rate
Optimal
ordering
Optimal
ordering

Multi-item + multiperiod
Nonperishable
Veinott (1965)

GA
GA

GA/SA

Budget

Storage

Quantity
discount

No
shortages

Lot-size

St, Se, D

Shortages

GA/LS

(Q ,r)

Lifetime

Extra
quantity

Resource
constraint

St: Stochastic; Se: Seasonal; D: Deterministic; F: Fixed; H: Heuristics; GA: Genetic Algorithm; LS: Local Search; R: Random; HGA: Hybrid Genetic Algorithm.

8192

I. Saracoglu et al. / Expert Systems with Applications 41 (2014) 81898202

pharmacies. This distributor wants to determine when to make an

order and how much to order to meet the customer needs in a way
that maximizes prot given certain budgetary constraints. Some
products require cold storage; therefore, storage area constraint
exists for these products. Further, manufacturing companies may
offer extra quantities as promotion according to the given order
quantity because of competition. In addition, the products have
specic lifetimes. As a result, the FIFO rule is applied, that is, the
product already on hand is sold prior to the sale of more recently
delivered products. Because demand varies by product and the
demand for some products is uncertain or seasonal, a multi-product multi-period (Q , r) model has to be applied. Accordingly, a specic order quantity (Qi) is given whenever the stock level for a
product i is at or below level ri in this system. The order arrives
at the storage with a delay of a certain lead time. If there is a promotion on an order quantity Qi, the quantity that is added to our
storage is then Qi + extra quantityi..
We make the following assumptions in expressing our multiproduct multi-period (Q , r) inventory problem mathematically:
 There are multiple products.
 There is only one warehouse.
 Demand is deterministic, seasonal, or variable according to
products.
 Lead time is xed and constant.
 There is a storage limit for products that require cold storage.
 A nite budget limitation is enforced by the bank.
 Backorders are not allowed.
 Shortages are allowed.
 Depending on the size of the order, extra quantities are supplied
by the manufacturers as promotions.
 The planning horizon is nite and planning is periodical.
 The manufacturers and customers both have payment terms,
which are different for each product.
 Lifetime is xed and constant, and changes according to the
product.
3.1. Integer linear programming model for multi-product multi-period
(Q , r) policy
The problem at hand is to formulate a model that maximizes
the total prot subject to constraints on budget, storage area, shelf
life, and various promotions. The following notations are used to
derive the ILP formulation of the proposed multi-product multiperiod (Q , r) continuous review inventory model.
3.1.1. Notations
The sets, indices, parameters, and decision variables used in the
model are given as follows:
Sets and indices
N
T
K
PC
Parameters
Qi,t,E
xi,t

number of products, where, i = 1, 2, . . .,

N.
number of planning periods, where,
t = 1, 2, . . ., T.
number of break points for extra
quantity, where, k = 1,2, . . . ,K.
set of products that need cold storage
environment.
extra order quantity for product i in
period t.
expected demand for product i in
period t.

mdi
li
spi
ppi
csi
sci
hi
Ie
Mi

Ni
wi
I1i,0
shfi
upper_ri
upper_Qi
Uk
Lk
EXi,k
B
storage
BigM
Decision variables
Ii,t
I1i,t
I2i,t
Qi,t
QSi
ri
Yi,t
Ai,t

Bi,t

Di,t,k

expected mean demand during the

lead time for product i.
lead time for product i.
sales price for product i.
unit purchasing price for product i.
ordering cost for product i, per order.
unit shortage cost for product i.
unit holding cost for product i, per
period.
interest rate per period.
the warehouses payment term offered
by manufacturer in unit time for
product i.
the customers payment term offered
by distributor in unit time for product i.
storage space required for product i.
initial inventory level for product i.
lifetime period for product i.
reorder point upper bound for product
i.
order quantity upper bound for
product i.
upper break point for quantity range k.
lower break point for quantity range k.
extra goods for quantity range k of
product i.
total available budget.
total storage capacity for all products.
big number.
ending inventory level for product i in
period t.
positive inventory level for product i in
period t.
shortage quantity of product i at the
end of period t.
order quantity for product i in period t.
order quantity for product i.
reorder point for product i.
binary variable: equal to one if an order
is placed in period t; otherwise, zero.
binary variable: equal to one if the
beginning inventory level of product i
is less than the reorder point in period
t, and zero if the beginning inventory
level is greater than the reorder point,
indicating whether an order should be
placed.
binary variable: equal to one if the
ending inventory level of product i in
period t is positive, and zero if the
inventory level is negative.
binary variable: equal to one if a
purchase is made in period t for
product i, and zero if no purchase is
made in period t for product i, with
quantity break k.

3.1.2. Objective function

Our aim is to maximize prot; therefore, the objective function
is the total of the sales revenue and money earned as a result of the
payment terms provided by the manufacturers, minus the total

8193

I. Saracoglu et al. / Expert Systems with Applications 41 (2014) 81898202

purchasing, holding, ordering, and shortage costs, and money lost

because of the payment terms provided to customers.
Consequently, the objective function of the multi-product,
multi-period (Q , r) policy model is given in the equation below:
Maximize Total Profit Expected sales rev enue
Earnings resulting from payment term prov ided by manufacturer
 Total cost  Loss resulting from payment term prov ided to customer
1
Total Cost Purchasing cost Holding cost Ordering cost
Shortage cost

MaxZ Q i ; r i

The components of the objective function can be calculated as the

following relations:
Expected sales revenue:

ETSR

T X
N
X

xi;t  spi 

t1 i1

T X
N
X

I2i;t  spi

t1 i1

The total sales revenue obtained during the planned T periods is the
product of the demand quantity and unit sales price. If the incoming
demand is not completely satised, sales loss will occur as a result
of the unmet demand. Expected sales revenue is calculated by
subtracting this sales loss from the sales revenue.
Earnings resulting from the payment term provided by
manufacturer:
N X
T
X
TPFS
Q i;t  ppi  Ie  M i

Manufacturing companies may provide distributors with convenient payment terms for purchased products. When the manufacturer company provides the distributor with the ease to make the
payment after the sales of the product, the distributor indicates
the interest income for not having to make the payment immediately. The earnings that the distributor will obtain during the
planning period as a result of the payment terms is calculated as
in Eq. (5).
Purchasing cost:

TPC

Q i;t  ppi

t1 i1

Each period, consideration is given to whether an order should be

placed or not and, if an order is placed, the total purchasing cost
that will accrue during the planning period is equal to the product
of the unit sales price and order quantity.
Loss resulting from the payment term provided to customer:
N X
T
X
TLPFC
xi;t  spi  Ie  Ni

i1 t1

A distributor may also provide customers with convenient payment

terms similar to those provided by the suppliers of the distributor.
In this case, the distributors loss is the product of the sales amount,
payment term, and interest.
Holding cost:

THC

N X
T
X
hi  I1i;t

i1 t1

The total holding cost (THC) during the planning period is the product of the available stock quantity at the end of each period and the
periodic holding cost.
Ordering cost:

Q i;t 6 BigM  Y i;t

TOC

i 1; . . . ; N; t 1; 2; . . . ; T

N X
T
X
Y i;t  csi
i1 t1

ETSC

N X
T
X
sci  I2i;t

11

i1 t1

If the demand is higher than the level of the reorder point in period
t, then there can be a shortage quantity, in which case, the expected
total shortage cost (ETSC) is the product of the expected shortage
quantity and the unit shortage cost.
3.1.3. Constraints
The constraints of the model can be grouped into six sets:
budget, storage area, lifetime, inventory balance equations, and
extra quantity. These constraints are explained in detail in the
ensuing subsections.
Inventory balance equations:

If
Ii;t1 6 r i

i 1; . . . ; N; t 1; . . . ; T

12

Then
Q i;t Q i;t;E  I1i;t I2i;t I1i;t1  xi;t 0
i 1; . . . ; N; t 1; . . . ; T

13

Else

i1 t1

T X
N
X

If Qi,t > 0, then an order has been placed. Constraint (9) is used to
determine whether an order has been placed or not. The total ordering cost (TOC) is the product of the number of orders given during
the total planning period and the ordering cost.
Expected Total Shortage cost:

9
10

I1i;t1  I1i;t I2i;t  xi;t 0

i 1; . . . ; N; t 1; . . . ; T

14

and
Q i;t 0 i 1; . . . ; N; t 1; . . . ; T

15

I1i;t1  r i 1 6 BigM  1  Ai;t

ri  I1i;t1 6 BigM  Ai;t

i 1; . . . ; N; t 1; . . . ; T 16

i 1; . . . ; N; t 1; . . . ; T

Q i;t Q i;t;E  I1i;t I2i;t I1i;t1  xi;t P BigM  1  Ai;t

i 1; . . . ; N; t 1; . . . ; T

17
18

Q i;t Q i;t;E  I1i;t I2i;t I1i;t1  xi;t 6 BigM  1  Ai;t

i 1; . . . ; N; t 1; . . . ; T

I1i;t1  I1i;t I2i;t  xi;t P BigM  1  Ai;t

i 1; . . . ; N; t 1; . . . ; T


I1i;t1  I1i;t I2i;t  xi;t 6 BigM  1  Ai;t
i 1; . . . ; N; t 1; . . . ; T
Q i;t 6 BigM  Ai;t

i 1; . . . ; N; t 1; . . . ; T

19
20
21
22

According to the relations given in Constraints (12)(15), if the

stock level at the beginning of a period is lower than the reorder
point implied by Constraint (12), an order quantity Qi,t is placed
and the inventory balance equation is enforced by Constraint (13).
According to Constraint (13), for product i, inventory level Ii,t at
the end of period t is found by adding order quantity Qi,t and extra
quantity Qi,t,E to inventory level Ii,t1 at the beginning of the period,
and subtracting expected demand xi,t. Conversely, if the stock level
at the beginning of a period is greater than the reorder point, the
inventory balance equation is provided by Constraint (14), and no
order is placed by Constraint (15). In order to model these relations
linearly, Constraints (16)(22) were developed for the proposed ILP
model.
Order quantity constraints:



Q i;t  QSi 6 BigM  1  Ai;t
i 1; . . . ; N; t 1; . . . ; T


Q i;t  QSi P BigM  1  Ai;t
i 1; . . . ; N; t 1; . . . ; T

23
24

Constraints (23) and (24) assert that the order quantities should be
equal for all the periods if an order has been placed, which is indicated by QSi for each product.

8194

I. Saracoglu et al. / Expert Systems with Applications 41 (2014) 81898202

Inventory level constraints:

I2i;t 6 BigM  Bi; t i 1; . . . ; N; t 1; . . . ; T

I1i;t 6 BigM  1  Bi; t i 1; . . . ; N; t 1; . . . ; T
Ii;t I1i;t  I2i;t ;

i 1; . . . ; N; t 1; . . . ; T

25
26
27

Constraints (25) and (26) dene positive and negative inventory as

I1i,t = max{Ii,t, 0} and I2i,t = max{Ii,t, 0}, respectively. In period t, the
inventory holding cost is (hiI1i,t), and the inventory shortage cost is
(sciI2i,t). Constraint (27) nds the on-hand inventory level or shortage quantity.
Storage space constraints:

X

wi
I1i;t1 Q i;t Q i;t;E 6 storage i 1; . . . ; N; t 1; . . . ; T
i2PC

28
This constraint ensures that the maximum inventory level is less
than or equal to the storage area capacity in each period. Because
some products in the storage need a cold environment, storage constraint has to be taken into account for such products.
Budget constraints:
N X
T
X
Q i;t  ppi 6 B

29

i1 t1

The product of the order quantity planned for the entire planning
period and the purchasing price should be below the budget. The
distributor should complete the purchasing process without
exceeding this credit limit.
Lifetime constraints:

I1i;t1 Q i;t Q i;t;E

6 shfi
mdi

i 1; . . . ; N; t 1; . . . ; T

30

Products have specic lifetimes; the products have to be destroyed

if these lifetimes expire. Therefore, the time obtained by dividing
the total inventory at the beginning of the period and the incoming
orders by the average demand should be shorter than the lifetime.
Extra quantity constraints:

Q i;t;E

8
EX i;1 ; L1 6 Q i;t < U 1
>
>
>
>
< EX i;2 ; L2 6 Q i;t < U 2

..
>
>
.
>
>
:
EX i;k ; Lk 6 Q i;t < U k

31

In an effort to stay ahead of the competition, manufacturing companies may send various extra quantities, EXi,k, of a product according
to the quantity purchased by the distributor. Constraint (31) shows
the extra quantities offered by a manufacturing company in return
for the quantity ordered. If the distributor places an order for i product between L1 and U1 in t period, s/he will obtain an extra quantity
of EXi,1. These products arrive in storage at no cost and are sold at
the normal sales price to the customer.

Q i;t  Lk P BigM  Di;t;k i 1; . . . ; N; t 1; . . . ; T ; k 1; . . . ; K



Q i;t  U k 1 6 BigM  1  Di;t;k
i 1; . . . ; N; t 1; . . . ; T; k 1; . . . ; K
K
X
Di;t;k 1 i 1; . . . ; N; t 1; . . . ; T

32
33
34

k1
K
X
Q i;t;E
EX i;k  Di;t;k i 1; . . . ; N; t 1; . . . ; T

35

i1

Q i 6 upper Q i i 1; . . . ; N

36

r i 6 upper ri i 1; . . . ; N

37

Ai;t ; Bi;t ; Y i;t ; Di;t;k 2 f0; 1g i 1; . . . ; N; t 1; . . . ; T; k 1; . . . ; K 38

r i ; QSi ; Q i;t ; Ii;t ; I1i;t ; I2i;t P 0; integer i 1; . . . ; N; t 1; . . . ; T
39

Constraints (32)(34) determine the quantity range within which

the given ordered quantity falls. Constraint (35) determines the
extra quantity in regard to the quantity purchased. Constraints
(36) and (37) assert that the ordered quantities and reorder points
should be less than or equal to the upper bound values specied for
each product. Constraints (38) and (39) ensure that the integer
restrictions for all the variables are used in the model.
4. Proposed genetic algorithm approach
In this section, we present our proposed GA solution approach,
which has an embedded local search for determining the optimal
or near optimal multi-product multi-period (Q , r) inventory policy.
GA is a stochastic search method proposed by Holland (1975) that
simulates the development process of biological systems based on
Darwins survival of the ttest principle. It is one of the most
extensively used algorithms in inventory problems with respect
to other heuristics. GA is a population-based adaptive search technique which is effective when the search space of the problem is
large. GA is one of the best solution approaches among other metaheuristics when large-scale problems are considered (Soleimani,
Seyyed-Esfahani, & Shirazi, 2013). It has been applied to our problem as well as to other inventory management related problems in
the literature (Gupta, Bhunia, & Goyal, 2009, Pasandideh et al.,
2011, Yang et al., 2012) successfully. Besides, GA facilitates an easy
solution representation for our problem requiring a large number
of variables and constraints and an easily applicable neighborhood
solution generation due to its population based search in comparison to single solution search based metaheuristics such as simulated annealing and tabu search.
Gen and Syarif (2005) proposed a hybridized tree-based GA to
solve multi-period production/distribution and inventory problems. They adopted a fuzzy logic controller to improve the performance of the algorithm, and considered multiple products with
independent demand, multiple shared resources, and multi-periods. Kannan, Sasikumar, and Devika (2010) proposed a GA to solve
the mixed-integer linear programming model for a multi-echelon,
multi-period, multi-product closed-loop supply chain network
model containing storage space and processing time restrictions
in order to minimize the total cost. Pasandideh, Niaki, and
Mousavi (2013) developed a random demand and limited storage
capacity multi-period multi-item inventory control problem with
the aim of determining the optimal order quantities of products
in different periods and minimizing the total inventory cost. They
considered the all-unit and incremental quantity discount options.
The modern eld of genetics provides a further explanation of
evolution and the natural selection involved in the survival of the
ttest. GAs start with an initial set of random solutions, called a
population, and then choose parents to reproduce.Each individual
in the population is called a chromosome. A simple GA that renders
good solutions in many practical problems is composed of three
operators: selection, crossover, and mutation. The efciency of
GAs depends on many parameters, such as the initial population,
the representation of individuals, the selection strategy, and the
recombination of crossover and mutation operators.
4.1. Chromosome representation
The rst and most important step in the GA approach is deciding on the chromosome structure. The GA approach begins with a
numerical demonstration of the decision variables and data
involved in the chromosome structure. In this paper, the decision
variables of the proposed model are reorder point, r, and order
quantity, Q. Matrix notation is used for the demonstration of a
chromosome. The rst and second rows of the matrix contain the

8195

I. Saracoglu et al. / Expert Systems with Applications 41 (2014) 81898202

order quantity (Qi) and reorder point (ri) variable, respectively

(i = 1, . . . ,N; the number of products). The columns of the matrix
show the products. Consequently, a chromosome comprises a 2
 N matrix, including two rows and as many columns as the
number of products. The structure of such a chromosome is
depicted in Fig. 1.
It is remarkable that binary coding chromosome structures are
predominantly used in lot-size scheduling problems in the literature, whereas real numbers are used in inventory problems. In
most cases, binary coding is not used to solve large decision
variable optimization problems (Gupta, Bhunia, & Goyal, 2007).
Further, in our problem, the chromosome structure is formed using
real values.
4.2. Initial population generation
GA starts a search with an initial population. It is well known
that the performance of metaheuristics is affected by the selection
of the initial solution (s). If the initial solution is sufcient, the possibility of reaching better solutions increases and convergence to
the near optimal or optimal solution will be faster. In order not
to depend solely on the initial population, a local search embedded
in the GA generates the neighborhood solutions of the best
chromosomes in the initial population and returns the best one
to the population. The initial population is determined randomly
so that the decision variables Q and r are between the lower and
upper limits. The lower limit is assumed to be zero for each product, whereas the upper limit is determined to be four times the calculated values of the
economic
EOQi and ri. For each
p
order quantity,
P
product, EOQ i 2csi  X i =hi , where X i Tt1 xi;t , and reorder
point is calculated as ri = mdili. Therefore, all the chromosomes
in the initial population are generated randomly as the size of
the population (pop_size).
4.3. Calculating the tness values
Each chromosome in the initial population is evaluated by its
tness value. Because the procedure used to select the parent chromosomes for reproduction depends on the tness values, it is
important to calculate them properly. The individuals with the
worst tness values die, whereas the individuals with the better
tness values survive. A penalty function is deployed to address
the problem constraints. Since the aim is to maximize prot, the
tness function is calculated by subtracting the penalty function
from the objective function as follows:

Fitness function objective function  penalty function

Consequently, an individual that has maximum prot in a
population has maximum tness value. The following penalty function, developed by Joines and Houck (1994), is used in this paper:

ev alx f x pt; x
m
X
b
di x
pt; x p/t

40
41

i1

where eval(x) is the evaluated tness function value, p(t, x) is the

variable penalty factor, t represents the generation of the GA, di(x)
is the penalty term for single constraint, and a and b are parameters
used to adjust the scale of the penalty value. pt = C  t, where C is a
constant. In this penalty function, it is essential that the parameter
values be chosen correctly. The recommended values are 1, 1 or 2,
and 0.5 for the parameters a, b, and C, respectively. However, we
used b = 5, a = 1, and C = 0.5 in our problem because the recommended parameter values did not result in high-quality solutions:
di(x) is calculated as the penalty term for each iteration. The

chromosome

gene
Fig. 1. Gene string and chromosome representation of 2  N matrix.

objective is to increase the penalty value as the number of iterations

increases in order not to choose infeasible solutions.
4.3.1. Calculation of the tness values for variable demand
The tness value calculation procedure for products with variable demand structure is different from those of the deterministic
structure. If the demand of a product is variable, J demand sets are
generated to represent the variability inherent in that demand. To
calculate the tness values of products with variable demand, the
corresponding (Q , r) values in the chromosome are put into the
tness function for each demand set and the tness value calculated
by taking the average of the tness values for each demand set.
4.4. Proposed local search
To avoid local optimum during the GA process, the process is
combined with different procedures. In this paper, a local search
approach, proposed by Zhao et al. (2012), is used to nd neighbor
solutions of the current (Q , r) solution by increasing and decreasing
Q and/or r in succession. Thus, neighbor solutions of (Q , r) are
obtained by changing Q and r; i.e., (Q , r) to (Q , r + 1), (Q , r  1),
(Q + 1, r), (Q  1, r), (Q + 1, r + 1), (Q  1, r  1), (Q + 1, r  1),
(Q  1, r + 1). If a chromosome that has a better tness value is
obtained in each iteration at the local search, that chromosome is
included to the population. This procedure is iterated as many
times as the determined number of iterations. The general structure of the GA with local search procedure is depicted in Fig. 2.
4.5. Selection
In this paper, roulette wheel selection (Gen & Cheng, 1997) is
utilized in the selection of parent chromosomes for reproduction.
The roulette wheel selection method is a probabilistic selection
approach that depends on the performance of the chromosomes.
The tness value, tnessvalue (vj), of each chromosome in the popP size
fitnessv alue v j ,
ulation and the total feasibility value, F pop
j1
for the population are calculated, where pop_size is the number of
chromosomes in the population. The probability value, pj = tnessvalue (vj)/F, for each chromosome and the cumulative probability
Ppop size
pj , are found. A random number (rand) is
value, qj j1
derived from the uniform [0, 1] range and the chromosome that
coincides with the range including that random number is selected
to produce new individuals.
4.6. Genetic operators: crossover
The design of the genetic operators has an important effect on
stabilizing the performance of GA. Genetic operators are grouped
into two types: crossover and mutation. Crossover performs the
mating process of the individuals that will produce new offspring
for the future new population. A mutation operator is used in the
process to randomly change the genetic value in the chromosome.
In the proposed GA approach, since our chromosome structure is a
2  N matrix, the crossover operator is applied in three different

8196

I. Saracoglu et al. / Expert Systems with Applications 41 (2014) 81898202

Fig. 2. General structure of the GA with local search.

ways. Information is given about the application of the proposed

crossover procedure related to our example problem as follows:
1. Changing just one row: If the random number we have derived in
the [0, 1] range is smaller than the rst crossover type rate Pc,1
that we have determined, then a new offspring is generated
by completely replacing the randomly selected row with the
row of the other parent. This crossover, Type 1, is depicted in
Fig. 3.
2. Changing the right or left hand side of both rows from the determined point: If the derived random number is smaller than
the determined second crossover type rate Pc,2, then both rows
are replaced together. The determination as to whether a crossover will be conducted from the right or left hand side of the

column is made with a 50% probability by selecting a random

number (crosspoint) in the [1, N1] range. This crossover, Type
2, is depicted in Fig. 4.
3. Crossover on the right or left hand side of a selected row: If the
derived random number is smaller than the determined third
crossover type rate, then a new offspring is obtained by replacing the right or left hand side of the selected row with a 50%
probability. This crossover, Type 3, is depicted in Fig. 5.
4.7. Genetic operators: mutation
This operator arbitrarily selects genes to counteract premature
convergence and to maintain enough diversity in the population.
Mutation is employed to maintain the diversity of the entire

8197

I. Saracoglu et al. / Expert Systems with Applications 41 (2014) 81898202

that virtually all children are different from their parents. The
genes selected for the mutation operator applied in this paper were
randomly changed between the lower and upper limits. The application of the mutation operator is discussed by Eiben and Smith
(2007) and Goldberg (1989).
4.8. Termination criteria
In general, two methods are used for terminating a GA. In the
rst of these methods, the algorithm is terminated after a predened number of generations (iter_num), whereas in the second
method, the algorithm is terminated when no further development
can be obtained in the solution after a predened number of
successive generations. In our proposed GA approach, we use the
former method.
4.9. Parameter optimization
Fig. 3. Only one row applied for crossover.

We stated above that generation number (iter_size), population

number (pop_size), crossover type rate (Pc,1, Pc,2), and mutation rate
(Pm) are the most important parameters in GA. In order to calibrate
the proposed GA, the Taguchi method is used in this paper to statistically design the experimental investigation. Traditional experimental design procedures are in essence too complicated and
difcult to use because a great deal of experimental work has to
be carried out as the number of process parameters increases. To
solve this problem, the Taguchi method uses a special orthogonal
arrays design to study the entire parameter space with only a small
number of experiments. We employ Taguchis orthogonal arrays,
the signal-to-noise (S/N) ratio, and analysis of variance (ANOVA)
to nd the appropriate levels and to analyze the effect of parameters on the objective values obtained. The S/N ratio indicates the
amount of variation present in the response variable. The aim is
to maximize the signal-to-noise ratio. Taguchi categorizes objective functions into three groups: the smaller-the-better type, the
larger-the-better type, and nominal-is-best type. Our objective
function is to maximize prots; therefore, we choose the largerthe-better type S/N ratio and calculate as follows:
Fig. 4. Crossover applied on the column.

S=N 10  log

n
1X
1
n i1 objectiv e function2i

!
41

where n is the replication number and objective functioni is the

result of experiments in replication i.
We dened four design factors that are important in GA performance for a three-product eight-period problem type: (1) iteration
size (iter_size), (2) population size (pop_size), (3) type rate (Pc,1,Pc,2),
and (4) mutation rate (Pm). The possible three levels are considered
for each design factor, as shown in Table 3. While 34 = 81 experiments should be made for four factors with three levels, Taguchi
analysis enables us to conduct only 27 experiments by choosing
the orthogonal array L27. By executing the GA 10 times for each
experiment, we obtained average, minimum, and maximum tness
values for each experiment.
On obtaining the results of the Taguchi experiment, we transformed them to S/N ratios. To explore the relative signicance of
individual factors in terms of their main effects on the objective
Table 3
Experimental factors and their levels.
Factors
Coding

Iteration
number
iter_num

Population
size
pop_size

Mutation
rate
Pm

Crossover
type
(Pc,1, Pc,2)

Level 1
Level 2
Level 3

200
500
1000

20
50
100

0.05
0.15
0.4

(0.25, 0.50)
(0.33, 0.66)
(0.50, 0.75)

Fig. 5. Crossover applied on both row and column.

population by changing individuals bit by bit with a small probability Pm 2 [0, 1]. Mutation prevents the algorithm from escaping
the local minima trap. Crossover and mutation operators ensure

8198

I. Saracoglu et al. / Expert Systems with Applications 41 (2014) 81898202

Table 4
ANOVA table of means.

Source

DF

Seq SS

Adj SS

Adj MS

Main effects
iter_num
pop_size
Pm
Pe
2-Way interactions
iter_num  pop_size
iter_num  P m
pop_size  P m
Pm  Pc
3-Way interactions
pop_size  Pm  Pc
Residual error
Total

4
1
1
1
1
4
1
1
1
1
1
1
17
26

1492494
455951
28267
965861
42416
522674
112564
137754
27378
244978
227173
227173
1164360
3406701

1802638
455951
36512
1183356
146676
316515
112564
137754
15521
66176
227173
227173
1164360

450660
455951
36512
1183356
146676
79129
112564
137754
15521
66176
227173
227173
68492

6.58
6.66
0.53
17.28
2.14
1.16
1.64
2.01
0.23
0.97
3.32
3.32

0.002
0.019a
0.475
0.001a
0.162
0.365
0.217
0.174
0.640
0.339
0.086
0.086

Factors effective at 95% condence level.

function, ANOVA was conducted: the results are shown in Table 4.

As can be seen in the table, factors with p 6 0.05 at 95% condence
level, i.e., iteration number and mutation rate, are effective on the
results. In addition, it can be seen that the interaction between the
four factors does not cause a signicant difference in the results.
Further, the most effective factor is the mutation rate that has
the highest F value. We can obtain information about which factor
is more effective with Taguchi analysis, as shown in Table 5.
Table 5 shows the average of the response characteristics (S/N
ratios, means) for each level of all factors. It includes the ranks
based on delta statistics comparing the relevant sizes of the effects.
Delta statistics is the difference between the greatest and the
smallest average values of each factor. S/N ratio and mean tables

Table 5
Response table for signal-to-noise ratios (larger is better).
Level

Iteration

Population

Mutation

Crossover

1
2
3
Delta
Rank

86.77
86.86
86.91
0.14
2

86.81
86.90
86.85
0.09
4

86.94
86.85
86.76
0.18
1

86.80
86.92
86.83
0.13
3

have the same rank values. The mutation parameter has the greatest effect on both the S/N ratio and the average. Iteration number
has the second- greatest effect, followed by crossover and population, respectively. As seen in Table 5, the levels with S/N ratio that
give the greatest value are the levels that are effective in the
solution. Finally, the parameter values that we should use are as
follows: {Pm = 0.05, iter_num = 1000, Pc,1, Pc,2 = (0.33, 0.66), and
pop_size = 50}.
5. Computational study
In this section, various test problems with varying numbers of
products and periods are solved and results compared with the
results of the ILP model in order to evaluate the performance of
the proposed GA. We used LINGO 9.0 software to solve the ILP
model and terminated the execution of the model after ve hours.
The optimality of the solutions in the rows marked with asterisks
was not conrmed because LINGO could not complete its search
for an optimal solution within the specied time limit. The GA
solutions were obtained from a MATLAB 7.11 implementation.
All test problems were solved on an Intel Core Duo PC with a
P8700@2.53 GHz CPU and 4.00 GB RAM.
Table 6 shows the total prot value and CPU time (seconds)
results of the ILP model and the proposed GA. The proposed GA

Table 6
Results obtained by ILP and the proposed GA.
Problem type

2P_4T
2P_8T
2P_10T
2P_11T
2P_12T
2P_14T
2P_15T
3P_4T
3P_8T
3P_10T
4P_4T
4P_8T
5P_4T
5P_8T
6P_4T
6P_8T
7P_4T
7P_8T
8P_4T
a

Product no

2
2
2
2
2
2
2
3
3
3
4
4
5
5
6
6
7
7
8

Period no

4
8
10
11
12
14
15
4
8
10
4
8
4
8
4
8
4
8
4

Interrupt after 5 h so the solution may not be optimal.

Total prot ($)

GAP (%)

ILP

Proposed GA-best

9792.86
19510.75
27606.49
30467.76
33870.36
40200.92a
43840.68a
10566.50
23531.93
29122.90a
10987.09
22621.60a
10984.86
24034.04a
11115.78
22237.50a
11146.60
21534.83a
11162.11a

9624.49
19221.80
27606.49
30466.99
33630.79
40200.92
43858.74
10566.50
23531.93
32004.85
10385.54
23029.78
10947.90
23828.11
11071.98
23852.91
11146.60
23493.55
11149.30

1.72
1.48
0.00
0.00
0.71
0.00
0.04
0.00
0.00
9.90
5.48
1.80
0.34
0.86
0.39
7.26
0.00
9.10
0.11

Computational time CPU (sec)

ILP

Proposed GA

5.00
69.00
756.00
1457.00
13698.00
18000.00
18000.00
26.00
17259.00
18000.00
54.00
18000.00
162.00
18000.00
596.00
18000.00
2812.00
18000.00
18000.00

22.80
29.87
32.51
35.27
35.45
39.00
40.73
36.44
41.82
53.77
44.45
57.43
55.31
83.67
68.17
98.73
83.64
111.51
104.92

GAP (%)

356.00
56.71
95.70
97.58
99.74
99.78
99.77
40.15
99.76
99.70
17.69
99.68
65.86
99.53
88.56
99.45
97.03
99.38
99.42

I. Saracoglu et al. / Expert Systems with Applications 41 (2014) 81898202

terminated when the iteration number reached 1000, and 20 replications were applied in order to nd the best value for the prot.
The table shows that better results can be obtained with GA in
much shorter computation times.
We calculated GAP (%) values for the total prot and solution
times using Eq. (42) in order to clearly compare the solution results
and times achieved by the ILP model and the GA approach. The
values computed are shown in Table 6.

GAP%

GA Solution  ILP Solution

 100
ILP Solution

42

While an optimal solution for a two products 12 periods [2P_12T]

problem could be obtained in approximately 3 h and 48 min by
the ILP model, a solution only 0.71% worse was reached in 35.45
CPU seconds by the proposed GA, incurring 99.74% less time. As
the number of products and periods increases, reaching an optimum solution with ILP becomes much more difcult. Consequently,
it may not be possible to reach even an integer feasible solution.
6. Case study of a pharmaceutical distributor in Turkey
The company addressed in this paper is a pharmaceutical distributor performing drug distribution with a single regional storage. Because the pharmaceutical distributor stores approximately
5000 products and distributes to 900 pharmacies, the products
are classied according to ABC analysis. The demand structure of
the products is decided on the basis of the percentage of the coefcient of variation V (standard deviation/mean demand) values
calculated for a given number of periods, as described by Taha
(2007). The products that have V values below 20% are considered
deterministic products, whereas those with V values greater
than 20% are considered within the context of variable products.
A quarter-based comparison is made for the products whether they
have a variable demand structure or a seasonal one. In the end,
nine products were chosen to represent the deterministic,
seasonal, and variable product groups from each ABC class, as
shown in Table 7.
The study has been initiated in 2012. The input values such as
initial stock, purchasing and sales prices, demand have been
obtained from the companys database. The sales data of the products for the last 4 years have been obtained from the Intercontinental Marketing Services [IMS] Health company in Turkey and
the variability of the demand structure of the products has been
examined. ANOVA analysis has been performed to examine
whether there is difference in the demand of products between
the periods. Inventory holding cost has been calculated using
warehouse rent, personnel maintenance, and utilities data of the
company along with the ratios given in Greene (1997). Ordering
cost includes the costs of purchasing department and that of the
processes of product purchasing, acceptance and control. Shortage
cost has been calculated as given in Nahmias (1993). The distributor may face customer loss if the order of a pharmacy cannot be

8199

met for ve times by the distributor. We tried to estimate the

stockout cost by using Type I service level as indicated in
Nahmias (1993). A one-year planning horizon is handled as 52
periods (weeks) in our study. If a customer is lost when it is
rejected for 5 times by the distributor throughout the planning
period, it is expected that the demand is met for 52  5 = 47 times.
Therefore, a is calculated as a = 47/52 = 0.90. This value corresponds to the F(r) value and as a result, stockout cost is computed
as in Eq. (43), where F(r) represents the average percentage of the
demands which are satised as they occur.

sci Q i hi =1  FrX i 

43

The stockout cost is calculated as sc2 =

(190)(3.45)/
[(1  0.90)(6228)] = $1.09 for Pro2.
The lifetime of a drug begins as soon as its production is complete and the expiration date is indicated on the package. The
amount of time that elapses during the transportation from manufacturer to the distributor is unknown. Pharmacies are generally
reluctant to buy products that only have a short lifetime left. Products with a lifetime equal to or less than one year are categorized
as short lifetime products. In this paper, the lifetimes of the drugs
are taken as 1/4 of their real lifetimes, thus guaranteeing that they
are sold before their lifetime expires. Pharmaceutical distributors
can obtain customer satisfaction initially by providing requested
products and through good communication and reliability. Further,
the convenient payment terms provided to the pharmacy is an
important factor in this selection. Because of the intense competition in the pharmaceutical industry, manufacturing companies
may offer extra quantities or various promotions according to the
purchased quantities. An extra quantity can be offered in particular
for equivalent and generic drugs. The extra quantities provided for
the selected nine drugs according to the ordered quantity range are
shown in Table 9. These promotions may vary for each product and
manufacturer.
The distributor has a yearly budget of $1,200,000 for the
selected nine products, with the planning horizon taken as one
year with weekly planning periods. Some products have to be
stored under cold storage conditions. Pro5 is an example of such
a drug. The storage capacity for this drug is taken as 1500 containers/period. Because the storage cost is higher for this type of product, inventory holding cost percentage is taken as 10% higher.
We generated demand sets for Pro3, Pro6, and Pro9 using probability distributions Weibull (66.5, 0.459), Beta (87, 0.702, 1.51),
and Normal (42.1, 30.2), respectively. The representation of the
variability of these product demands by the given probability functions were already conrmed via the statistical test.
Next, we determined the lower and upper bound values for our
(Q , r) decision variables. Our chromosome structure is represented
by a 2  9 (Q , r) decision variables matrix. The lower (lower_Qi,
lower_ri) and upper bound values (upper_Qi, upper_ri) are listed in
Table 10.

Table 7
Products selected to represent the various groups along with their descriptions.
Product name

ABC class

Demand type

Product description

Pro1
Pro2
Pro3
Pro4
Pro5
Pro6
Pro7
Pro8
Pro9

A
A
A
B
B
B
C
C
C

Deterministic
Seasonal
Variable
Deterministic
Seasonal
Variable
Deterministic
Seasonal
Variable

Helps treat asthma

Treats many different infections caused by bacteria
Relaxes muscles and increases blood ow to particular areas of the body
Antibiotics for surgery infection
Breaks disulde bonds in mucus and liquees it
Inhibits gastric acid secretion
Treats chronically elevated levels of iron in the blood
Alleviates itching pain related to insect sting and simple skin irritations
Alleviates nausea

8200

I. Saracoglu et al. / Expert Systems with Applications 41 (2014) 81898202

Table 8
Input data for the products.
Product i

Lead time (li) (week)

Economic order quantity (EOQi) (unit)
Purchasing cost (ppi)($)
Sales price (spi)($)
Inventory holding cost (hi)($)
Ordering cost (csi)($)
Shortage cost (sci)($)
Interest rate (Ie)(%)
Payment term for distributor (Mi) (week)
Payment term for pharmacy (Ni)(week)
Demand type (Di)
Shelf life (shfi) (week)
Storage limit (wi) (unit)
Initial invenory (I0i) (unit)

1
102
59.70
63.88
0.21
10
2.14
15
12
8
0
12
0
156

1
206
17.26
18.64
0.06
10
1.09
15
16
12
1
12
0
186

1
132
55.39
59.27
0.20
10
1.64
15
17
10
1
24
0
412

1
462
3.40
3.70
0.01
10
0.47
15
3
3
0
16
0
187

1
159
8.05
8.80
0.03
10
0.66
15
12
8
1
10
1500
753

1
95
17.30
18.70
0.06
10
2.27
15
17
12
1
36
0
26

1
5
1209.79
1233.99
4.31
10
43.14
15
12
8
0
12
0
2

1
294
3.40
3.70
0.01
10
0.74
15
17
12
1
36
0
49

1
308
2.72
3.00
0.01
10
0.71
15
8
4
1
36
0
43

Table 9
Extra promotional quantities offered with some products.
Pro2

Pro3

Pro4

Pro5

Pro6

Lower
limits

Upper
limits

Extra
quantities

Lower
limits

Upper
limits

Extra
quantities

Lower
limits

Upper
limits

Extra
quantities

Lower
limits

Upper
limits

Extra
quantities

Lower
limits

Upper
limits

Extra
quantities

10
50
100
150
200
250
400

49
99
149
199
249
399
599

0
4
5
10
15
18
39

0
5
10
20
50
250
500

4
9
19
49
249
499
999

0
1
3
8
30
250
500

0
5
10
50
100
250
500

4
9
49
99
249
499
749

0
5
15
90
240
750
1500

0
5
10
50
100
250
500

4
9
49
99
249
499
and
more

0
0
5
10
100
100
200

0
5
9
16
26
63
96

4
8
15
25
62
95
160

0
1
2
4
6
10
24

600
1200

1199
and
more

119
500

1000
1500

1499
2999

1000
1500

750
1000

999
1999

2250
3000

161
242

241
338

34
54

3000

and
more

3000

2000

2499

6000

339

and
more

78

2500

and
more

7500

With a $1,200,000 budget and all other constraints, the total

prot that the pharmaceutical distributor will obtain at the end of
the planning horizon is estimated as $334,642.19. This prot
level can be reached by placing orders of Q i = {241, 600, 500,
121, 100, 96, 6, 133, 207} order quantities at r i f101; 203;
392; 182; 237; 30; 5; 59; 73g inventory levels. Ten independent
replications were applied for the proposed GA approach; the details
of the best solution obtained are presented in Table 11. The results
were obtained at 90% service level and with the stock holding,

Table 10
The lower and upper bounds for the (Q, r) decision variables.
Products

Pro1

Pro2

Pro3

Pro4

Pro5

Pro6

Pro7

Pro8

Pro9

lower_Qi
upper_Qi
lower_ri
upper_ri

0
400
0
400

0
760
0
480

0
480
0
640

0
1700
0
480

0
1212
0
964

0
360
0
100

0
20
0
20

0
1080
0
192

0
1120
0
180

Table 11
Solution results for the Pharmaceutical Distributor.
Products

Pro1

Pro2

Pro3

Pro4

Pro5

Pro6

Pro7

Pro8

Pro9

Total

Q
r
Expected total sales revenue
(ETSR) ($)
Total earning from payment
term TPFS ($)
Total purchasing cost (TPC) ($)
Total losing from payment term
(TLPFC) ($)
Total holding cost (THC) ($)
Total ordering cost (TOC) ($)
Expected total shortage cost
(ETSC) ($)
Expected total shortage
quantity (ETSQ) (Unit)
Expected total prot (ETPR) ($)

241
101
332170.8

600
203
127178.72

500
392
383785.40

121
182
23125.44

100
237
134714.72

96
30
23087.81

6
5
320834.80

133
59
8230.65

207
73
6919.84

1360048.20

10514.53

4564.92

5541.76

60.84

668.43

809.14

13577.47

360.89

150.21

36248.23

302141.70
7706.36

98382.00
4425.81

191095.5
11129.77

6993.80
201.19

57624.00
1562.69

17438.40
803.45

312125.82
13956.31

7777.84
286.42

6474.96
80.27

1000054.02
40152.30

216.43
105
53.13

568.92
430
0

43.39
172
151.18

3182.3

38.5

71778.93

5280.54

7331.21

1309.56
210
0
0
31317.70

1517.20
95
389.574
282.3
26934.04

8155.28
69
3089.644
796.3
175787.97

100.63
170
0
0
15720.66

497.0025
420
3500.53

487.7
160.69

68.19
115
1.224

12476.62
1786.00
7185.28

5.1

4792.2

330.41

334642.19

I. Saracoglu et al. / Expert Systems with Applications 41 (2014) 81898202

.
.
Fig. 6. Prot versus inventory holding cost % change.

.
.

In accordance with actual company values, the inventory holding cost was taken as 18.51% of one unit purchasing cost. For this
reason, we researched how our expected maximum prot value
will change for a 40% change in the inventory holding cost. If
the inventory holding cost decreases by 40%, then the expected
maximum prot value increases by 1.47%. If the inventory holding
cost increases by 40%, then the expected maximum prot value is
observed to decrease by 4.37%. The effect of inventory holding cost
change on the expected maximum prot value is depicted in Fig. 6.
We estimated ordering cost at $10 per order. When the ordering
cost decreases by 50%, the change in the expected maximum prot
value is only 0.69%. Further, when the ordering cost increases by
50%, there is a 0.52% decrease in the prot value. The change for
the ordering cost and the expected maximum prot value is shown
in Fig. 7.
In accordance with actual company values, we assumed stockout cost as a loss from the prot. The obtained stockout cost corresponded to 90% customer service level (CSL). If CSL is chosen as
99%, the prot value decreases by 11.82%, whereas if CSL is chosen
as 99.5%, the prot value decreases by 20.07%. Thus, we can assert
that CSL has an important effect on the prot value. Stock holding
cost increases as the service level increases. The effect of service
level change on the expected maximum prot value is shown in
Fig. 8.
7. Conclusions

Fig. 7. Prot versus ordering cost % change.

.
.
.

8201

Fig. 8. Percentage change in prot for customer service level.

ordering, and stockout cost data given in Table 8. The computational time for an average replication was 2,737.07 s.
According to the current policy adopted by the company, they
examine the stocks one by one every week and place an order if
their budget is available considering the promotions and sales
within the last three months. Moreover, the calculation of reorder
point and xed-order quantity in the sector has been determined
to be very useful as a result of the negotiations. We are in constant
cooperation with the company and in case the purchases suggested
by us are applied, the following benets can be observed compared
to their own purchases. For example; for Pro2, whereas the prot
of the company at the end of its purchases is 32,456.81$, it would
be possible to obtain a prot of 26,934.04$ if the purchase policy
suggested by us were used.

The relevant literature review shows that the (Q , r) inventory

model has been only used as single period in multi-product problems and for nonperishable products. It is also found out that, for
multi-product multi-period case apart from the continuous review
inventory models, it is primarily dealt with lot-sizing issues in production planning and there is no study for perishable items as well.
In this paper, we formulated a novel approach for multi-product,
multi-period (Q , r) inventory models, with the objective of maximizing the prot under constraints such as storage area, budget,
shelf life, and various promotions. As the number of products and
periods increase, we found that an optimal solution cannot be
reached with ILP in a reasonable computational time. For this reason, we proposed a GA solution approach since the problem of the
pharmaceutical distributor case studied is a larger sized problem.
To determine the efcacy of the proposed GA, we compared the
solution results obtained for problem instances of various sizes to
those obtained from the ILP model. The analysis results showed
that the proposed GA approach can produce efcient and effective
solutions. Finally, we gave real results for our case study.
In future research, this model can further be generalized to
include safety stock in the chromosome structure proposed for
the GA solution approach. In addition, we will generalize the model
to allow partial backlogging and/or not allow shortages to increase
the service level. We also plan to handle supply chain management
problems including multi-suppliers and multi-customers in the
model. Finally, the performance of the proposed GA approach can
be examined with other metaheuristic methods and besides, other
inventory control policies such as (s, S)- and (Q, r, T) can be applied
on the problem.

6.1. Sensitivity analysis

References

In this section, because inventory holding cost, ordering cost,

and stockout cost are difcult to decide, sensitivity analysis is
performed in order to see the effect of these costs on the prot
value. The sensitivity analysis is shown by increasing or decreasing
cost parameters, taking one at a time and keeping the others at their
own values. In each case, the best value of optimal prot is found.

Ang, M., Song, J.-S., Wang, M., & Zhang, H. (2013). On properties of discrete (r, Q) and
(s, T) inventory systems. European Journal of Operational Research, 229, 95105.
Axster, S. (1993). Exact and approximate evaluation of batch-ordering policies for
two-level inventory systems. Operations Research, 41(9), 777785.
Axster, S. (2007). A heuristic for triggering emergency orders in an inventory
system. European Journal of Operational Research, 176, 880891.
Berk, E., & Gurler, U. (2008). Analysis of the (Q , r) inventory model for perishables
with positive lead time and lost sales. Operations Research, 56(5), 12381246.

8202

I. Saracoglu et al. / Expert Systems with Applications 41 (2014) 81898202

Betts, J. M., & Johnston, R. B. (2005). Determining the optimal constrained multiitem (Q , r) inventory policy by maximizing risk-adjusted prot. IMA Journal of
Management Mathematics, 16(4), 317338.
Bijvank, M., & Vis, I. (2012). Inventory control for point-of-use locations in hospitals.
Journal of the Operational Research Society, 63, 497510.
Eiben, A. E., & Smith, J. E. (2007). Introduction to evolutionary computing (2nd ed.).
Berlin Heidelberg: Springer-Verlag.
Federgruen, A., & Zheng, Y. (1992). An efcient algorithm for computing an optimal
(r, Q) policy in continuous review stochastic inventory systems. Operations
Research, 40(4), 808813.
Gardner, E. S. (1983). Approximate decision rules for continuous review inventory
systems. Naval Research Logistics Quarterly, 30, 5968.
Garman, M. B. (1976). Multi-product economic order quantity analysis under
minimum inventory valuation constraints. Operational Research Quarterly, 27(4),
983989. Pergamon Press.
Gen, M., & Cheng, R. (1997). Genetic algorithms and engineering design. John Wiley
and Sons.
Gen, M., & Syarif, A. (2005). Hybrid genetic algorithm for multi-time period
production/distribution planning. Computers & Industrial Engineering, 48,
799809.
Goldberg, D. E. (1989). Genetic algorithms in search optimization and machine
learning. Addison Wesley Longman, Inc.
Goyal, S. K., & Giri, B. C. (2001). Recent trends in modeling of deteriorating
inventory. European Journal of Operational Research, 134, 116.
Greene, J. H. (1997). Production & inventory control handbook (3rd ed.). A division of
the McGraw-Hill Companies.
Gupta, R. K., Bhunia, A. K., & Goyal, S. K. (2007). An application of genetic algorithm
in a marketing oriented inventory model with interval valued inventory costs
and three-component demand rate dependent on displayed stock level. Applied
Mathematics and Computation, 192, 466478.
Gupta, R. K., Bhunia, A. K., & Goyal, S. K. (2009). An application of genetic algorithm
in solving an inventory model with advance payment and interval valued
inventory costs. Mathematical and Computer Modelling, 49, 893905.
Hadley, G., & Whitin, T. M. (1963). Analysis of inventory systems. Englewood Cliffs,
New York: Prentice-Hall.
Hajiaghaei-Keshteli, M., Sajadifar, S. M., & Haji, R. (2011). Determination of the
economical policy of a three-echelon inventory system with (r, Q) ordering
policy and information sharing. International Journal of Advanced Manufacturing
Technology, 55, 831841.
Harris, F. W. (1913). How many parts to make at once? Factory. The Magazine of
Management, 10, 135152.
Holland, J. (1975). Adaptation in natural and articial systems. Ann Arbor: University
of Michigan Press.
Hopp, W. J., Spearman, M. L., & Zhang, R. Q. (1997). Easily implementable inventory
control policies. Operations Research, 45, 327340.
Jeddi, B. G., Shultes, B. C., & Haji, R. (2004). A multi-product continuous review
inventory system with stochastic demand, backorder, and a budget constraint.
European Journal of Operational Research, 158, 456469.
Joines, J. A., & Houck, C. R. (1994). On the use of non-stationary penalty functions to
solve nonlinear constrained optimization problems with GAs. Fogel, 132,
579584.
Kallrath, J., & Maindl, T. I. (2006). Real optimization with SAP APO. Berlin Heidelberg:
Springer-Verlag.
Kannan, G., Sasikumar, P., & Devika, K. (2010). A genetic algorithm approach for
solving a closed loop supply chain model: A case of battery recycling. Applied
Mathematical Modelling, 34, 655670.
Kim, J. S., & Benton, W. C. (1995). Lot size dependent lead times in a (Q , r) inventory
system. International Journal of Production Research, 33, 4158.
Kundu, A., & Chakrabarti, T. (2012). A multi-product continuous review inventory
system in stochastic environment with budget constraint. Optimization Letters,
6, 299313.

Lapierre, S. D., & Ruiz, A. B. (2007). Scheduling logistic activities to improve hospital
supply systems. Computers & Operations Research, 34, 624641.
Lau, AH.-L., & Lau, H-S. (2008). An improved (Q , r) formulation when the stockout
cost is incurred on a per-stockout basis. International Journal of Production
Economics, 111, 421434.
Louly, M-A., & Dolgui, A. (2013). Optimal MRP parameters for a single item
inventory with random replenishment lead time, POQ policy and service level
constraint. International Journal of Production Economics, 143, 3540.
Mandal, S., Maity, A. K., Maity, K., Mondal, S., & Maiti, M. (2011). Multi-item
multi-period optimal production problem with variable preparation
time in fuzzy stochastic environment. Applied Mathematical Modelling, 35,
43414353.
Mattsson, S- A. (2010). Inventory control in environments with seasonal demand.
Operations Management Research, 3, 138145.
Mousavi, S. M., Hajipour, V., Niaki, S. T. A., & Alikar, N. (2013). Optimizing multiitem multi-period inventory control system with discounted cash ow and
ination: Two calibrated meta-heuristic algorithms. Applied Mathematical
Modelling, 37, 22412256.
Nahmias, S. (1982). Perishable inventory theory, a review. Operations Research,
30(4), 680708.
Nahmias, S. (1993). Production and operations analysis (2nd ed.). Richard D. Irwin,
Inc.
Nahmias, S. (2011). Perishable inventory systems. Springer Science+ Business Media.
Pasandideh, S. H. R., Niaki, S. T. A., & Nia, A. R. (2011). A genetic algorithm for vendor
managed inventory control system of multi-product multi-constraint economic
order quantity model. Expert Systems with Applications, 38, 27082716.
Pasandideh, S. H. R., Niaki, S. T. A., & Mousavi, S. M. (2013). Two metaheuristics to
solve a multi-item multiperiod inventory control problem under storage
constraint and discounts. International Journal of Advanced Manufacturing
Technology, 69, 16711684.
Pasandideh, S. H. R., Niaki, S. T. A., & Tokhmehchi, N. (2011). A parameter-tuned
genetic algorithm to optimize two-echelon continuous review inventory
systems. Expert Systems with Applications, 38, 1170811714.
Rafaat, F. (1991). Survey of literature on continuously deteriorating inventory
models. The Journal of the Operational Research Society, 42(1), 2737.
Soleimani, H., Seyyed-Esfahani, M., & Shirazi, M. A. (2013). Designing and planning a
multi-echelon multi-period multi-product closed-loop supply chain utilizing
genetic algorithm. International Journal of Advanced Manufacturing Technology,
68, 917931.
Taha, H. A. (2007). Operations research an introduction (8th ed.). Pearson Education,
Inc.
Van Zyl, G. J. J. (1964). Inventory control for perishable commodities, unpublished
(Ph.D. dissertation). University of North Carolina.
Veinott, A. (1965). Optimal policy for a multi-product, dynamic, non-stationary
inventory problem. Management Science, 12, 206304.
Wang, T.-Y., & Hu, J-M. (2010). Heuristic method on solving an inventory model for
products with optional components under stochastic payment and budget
constraints. Expert Systems with Applications, 37, 25882598.
Yang, W., Chan, F. T. S., & Kumar, V. (2012). Optimizing replenishment polices using
genetic algorithm for single-warehouse multi-retailer system. Expert Systems
with Applications, 39, 30813086.
Zhang, S., & Eksioglu, S. D. (2009). Economic lot-sizing problem with multimode
replenishment and perishable inventory. In Proceedings of the 2009 industrial
engineering research conference.
Zhao, X., Fan, F., Liu, X., & Xie, J. (2007). Storage-space capacitated inventory system
with (r, Q) policies. Operations Research, 55(5), 854865.
Zhao, X., Qiu, M., Xie, J., & He, Q. (2012). Computing (r, Q) policy for an inventory
system with limited sharable resource. Computers and Operations Research, 39,
23682379.
Zheng, Y.-S. (1992). On properties of stochastic inventory systems. Management
Science, 38(1), 87103.