Available online at www.ispacs.

com/conferences/cjac
Volume 1, Year 2012, Article ID cjac-001-028, Pages 242-251
doi:10.5899/2012/cjac-001-028
Conference Article

The First Regional Conference on the Advanced Mathematics and Its Applications
29 FEB-1 MAR 2012, Mobarakeh Branch, Islamic Azad University, Mobarakeh, Iran.

ABC inventory classication withmultiple-criteria using

weighted non-linear programming
Z. Eslaminasab1*, T. Dokoohaki1
1

Department of Mathematics, Islamic Azad University, Khorasgan Branch, Isfahan

Copyright 2012 Z. Eslaminasab and T. Dokoohaki. This is an open access article distributed under the Creative Commons
Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work
is properly cited.

Abstract:
To have an ecient control of a huge amount of inventory items, traditional approach is
toclassify the inventory into dierent groups. Dierent inventory control policies can
thenapplied to dierent groups. The well-known ABC classication is simple to
understandand easy to use. In proposed method, the performance matrix is normalized
rst,then theideal alternative is introduced. The appropriate weights are obtained into
multiple criteriaby a non-linear programming. In next stage, alternatives are evaluated
based on DistanceBased Approach (DBA)method and classied. Finally, the proposed
method is comparedwith another methods by an example.
Keywords: ABC inventory classication; Distance Based Approach; Multiple criteria
analysis
1. Introduction
Inventory classication using ABC analysis is one of the most widely employed
technique in organization. Dierent inventory policies can the applied to dierent
groups.ABC analysis is a well-known and practical classication based on the pareto
principle.For example, group A inventory items are those making about 70% of
companys business but only taking up 10% of inventory. They are critical to the
functioning of the com-pany.Group B inventory items are those presenting about 20% of
companys business and taking about 20% of inventory. Group C items are those
representing 10% of company business but taking up about 70% of inventory. ABC
analysis is simple to understand and easy to use.However, traditional ABC analysis is
based on only single measurement such as annual dollar usage.It has been recognized
that other criteria,such as inventory cost, part criticality, lead time, commonality, obsolescence, substitutability, number of requests for the item in a year, scarcity, durability,

* Corresponding author. Email addresses: z.esb.math@gmail.com, dokoohakitahereh@yahoo.com

acta computare Volume 1, Year 2012, 242-251

243

substitutability, reparability, order size requirement, stock-ability, demand distribution,

and stock-out penalty cost are also important in inventory classication. Research literature on decision tools for multi-criteria inventory classication (MCIC) has been
developed in past 20 years.
Complex computational tools are needed for multi-criteria ABC classication. Flores et
al. [2] provide a matrix-based methodology. A joint criteria matrix is developed in the
case of two criteria. However, the methodology is relatively dicult to use when more
criteria have to be considered. Several multiple criteria decision-making (MCDM) tools
have also been employed for the purpose. Cohen and Ernst [3] and Ernst and Cohen [4]
have used cluster analysis to group similar items. The analytic hierarchy process (AHP)
[5] has been employed in many MCIC studies [2, 6, 7]. When AHP is used, the general
idea is to derive a single scalar measure of importance of inventory items by subjectively
rating the criteria and/or the inventory items [2, 8]. The single most important issue
associated with AHP- based studies is the subjectivity involved in the analysis. Heuristic
approaches based on articial intelligence, such as genetic algorithms [8] and articial
neural networks [9], have also been applied to address the MCIC problem. Clearly, these
approaches are heuris- tics and need not provide optimal solutions at all environments.
A very recent attempt by ramanathan [10] is to develop a weighted linear optimization
model to the problem. The basic concept of model in [10] is closely similar to the
concept of data envelopment analysis (DEA). The model rst converts all criteria
measurement into a scalar which is a weighted sum of measures under individual
criteria. To avoid the subjectively on the weight assignment, the weights are generated
by a DEA-like linea optimization. The clas- sication is then performed by grouping the
items based on the scores generated. however, a linear optimization is required for each
item. The processing time can be very long when the number of inventory items is large
in scale of thousands of items in inventory. Zhu and fan [11]present an extended version
of the Ramanathans model by incorporating some balancing features for MCIC. Zhu and
fan model, here after ZF-model, uses two sets of weights that are most favorable and
least favorable for each item. Ng [12] propose a simple model for MCIC. The model
converts all criteria measures of an inventory item into a scalar score. With proper
transformation, Ng obtains the scores of inventory items without a linear optimizer. The
Ng-model is exible as it could easily integrate additional information from decision
makers for inventory classication. But, Ng-model leads to a situation where the score of
each item is independent of the weights obtained from the model. That is, the weights do
not have any role for determining total score of each item and this may lead to a
situation where an item is inappropriately classied.In this article the performance
matrix is normalized based on desirable and un desirable criteria, at rst.The
appropriate weights are obtained into multiple criteria by a non-linear programming. In
next stage, alternatives are evaluated based on Distance Based Approach (DBA) method.
The mathematical formulation is presented in Section 2. An illustration is provided in
Section 3 with comparisons to the result from those in the literature. Short conclusions
are given in Section 4.
2. The proposed model
Consider an inventory with n items and these items are to be classied based on m
riteria. In particular, let the performance of ith inventory item in terms of the jth criteria

244

acta computare Volume 1, Year 2012, 242-251

e denoted as yij (i = 1, 2, ..., n; j = 1, 2, ...,m). We evaluate an item m (m = 1, 2, ...,M) by

converting multiple measures under all criteria into a single score. A common scale for
all measures is also an important issue. A particular criterion measure, in a large scale,
may always dominate the score. For this, we propose normalizing all measures yij into a
0-1 scale. We denote all transformed measures as xij . In order to transform the
performance ratings, the performance ratings are normalized into the range of [0, 1] by
the following equations [1].
(i)

The larger the better type:

(ii)

The smaller the better type:

*
*

(2.1)

+
+

Let Xj*= max {xij ; i=1, 2,..., n} denote the largest value that appears in the jth criterion.
Then I = ( X1*, X2*,, Xm* ) is empirically the ideal item because it has the best performance
in every criterion.
consider the following model :
2

*
1

)+

12
where Wj (j = 1, 2, ...,m) is the importance associated with the jth criterion and c
obtained
from the following model :

12
where is a small positive quantity imposed to restrict any criterion from being ignored.
when we have n items and m criterion, the whole set of items can be represented by the
following matrix [13] :

acta computare Volume 1, Year 2012, 242-251

245

11
21
[

12
22

1
2

that xij (i = 1, 2, ..., n; j = 1, 2, ...,m) denote the performance of ith inventory item in terms
of the jth criteria.
Thus in this matrix, a vector in an n-dimensional space represents every inventory
items.
In order to ease process and in the same time to eliminate the inuence of dierent units
of measurement, the matrix is standardized using z formula as :

where
i = 1,2,..., n.
m = Number of dierent criterion.
n = Number of inventory items.
xij = Indicator value for inventory item i for criteria j.
sj = Standard deviation of criterion j.
IN this study the optimal state of the objective is represented by the optimum inventory
item, the OPTIMAL. The vector I = ( X 1*, X2*,, Xm* ) is a set of optimum simultaneous
criterion. The vector I is called the optimal point. For practical purpose the optimal good
value for criterion is dened as the best value within the range of criteria. The standard
matrix is given as :

where

246

acta computare Volume 1, Year 2012, 242-251

The next step is to obtain the dierence or distance from each item to the optimal point
by subtracting each element of optimal by correspondence element in the item set. This
results in another interim matrix:

where
*

The next step is to introduce the aggregated performance weights for each selection
criteria.
If the aggregated performance weight for any selection criteria j is denoted by wj then
this will be result in another interim matrix as given:
(
(

)
)

(
(

)
)

(
(

)
)

Finally the Euclidean Composite Distance, CD, between each item inventory to the
optimal state, is derived from the following formula:

*(

where the weights are obtained from (2.3) and (2.4) And the items are ranked based on
the minimal distance to the optimal point.
3.Illustrative example
Following [12, 11] let us consider three criteria: Annual Dollar Usage (ADU), Average
Unit Cost (AUC) and Lead Time (LT) for inventory classication. All the criteria are
positive related to the score of the inventory items. An inventory with 47 items and
measurement of performance under each of the criteria considered are shown in Table
1. This table also shows the maximal and minimal measures under each criteria as well
as transformed

acta computare Volume 1, Year 2012, 242-251

247

measures in a scale of 0-1 as suggested in Section 2.

For comparison purpose, we maintain the same distribution of class A, B and C items as
in literature studies [10, 11], i.e. 10 class A, 14 class B and 23 class C. Table 2 shows the
classication based on our proposed model.The classication with the three criteria by
ZF-model, Ng-model and traditional ABC analysis using annual dollar usage are listed in
this table as well. As shown in Table 2, the ABC classication using our approach
provides dierent results compared with the other methods. Comparing to traditional
ABC analysis based on only annual dollar usage, only 25 out of 47 items are kept in the
same classes when ABC classication using proposed model with multi-criteria. seven
out of the ten class A items in traditional ABC classication is still classied as class A
items when multiple criteria is considered in proposed model.
The other two (S2, S9 and S10) are re-classied as class C using our model. For the 14
class B items are matched in both models, and for class C, 16 out of 23 items are
classified as class C items in both models.

248

acta computare Volume 1, Year 2012, 242-251

Table 1
Source and transformed measures of items under criteria
Item

ADU

AUC

LT

ADU
(transformed)

AUC
(transformed)

LT
(transformed)

S1
S2
S3
S4
S5
S6
S7
S8
S9
S10
S11
S12
S13
S14
S15
S16
S17
S18
S19
S20
S21
S22
S23
S24
S25
S26
S27
S28
S29
S30
S31
S32
S33
S34
S35
S36
S37
S38
S39
S40
S41
S42
S43
S44
S45
S46
S47

5840.64
5670
5037.12
4769.56
3478.8
2936.67
2820
2640
2423.52
2407.5
1075.2
1043.5
1038
883.2
854.4
810
703.68
594
570
467.6
463.6
455
432.5
398.4
370.5
338.4
336.12
313.6
268.68
224
216
212.08
197.92
190.89
181.8
163.28
150
134.8
119.2
103.36
79.2
75.4
59.78
48.3
34.4
28.8
25.38

49.92
210
23.76
27.73
57.98
31.24
28.2
55
73.44
160.5
5.12
20.87
86.5
11 0.4
71.2
45
14.66
49.5
47.5
58.45
24.4
65
86.5
33.2
37.05
33.84
84.03
78.4
134.34
56
72
53.02
49.48
7.07
60.6
40.
30
67.4
59.6
51.68
19.8
37.7
29.89
48.3
34.4
28.8
8.46

2
5
4
1
3
3
3
4
6
4
2
5
7
5
3
3
4
6
5
4
4
4
4
3
1
3
1
6
7
1
5
2
5
7
3
3
5
3
5
6
2
2
5
3
7
3
5

1.0000
0.9707
0.8619
0.8159
0.5939
0.5007
0.4806
0.4497
0.4124
0.4097
0.1806
0.1751
0.1742
0.1476
0.1426
0.1350
0.1167
0.0978
0.0937
0.0761
0.0754
0.0739
0.0701
0.0642
0.0594
0.0539
0.0535
0.0496
0.0419
0.0342
0.0328
0.0322
0.0297
0.0285
0.0269
0.0238
0.0215
0.0189
0.0162
0.0135
0.0093
0.0087
0.0060
0.0040
0.0016
0.0006
0.0000

0.7813
0.0000
0.9090
0.8896
0.7419
0.8725
0.8873
0.7565
0.6665
0.2416
1.0000
0.9231
0.6027
0.4861
0.6774
0.8053
0.9534
0.7833
0.7931
0.7397
0.9058
0.7077
0.6027
0.8629
0.8441
0.8598
0.6148
0.6423
0.3692
0.7515
0.6735
0.7662
0.7834
0.9904
0.7292
0.8257
0.8785
0.6960
0.7340
0.7727
0.9283
0.8409
0.8790
0.7892
0.8570
0.8844
0.9836

0.8333
0.3333
0.5000
1.0000
0.6666
0.6666
0.6666
0.5000
0.1666
0.5000
0.8333
0.3333
0.0000
0.3333
0.6666
0.6666
0.5000
0.1666
0.3333
0.5000
0.5000
0.5000
0.5000
0.6666
1.0000
0.6666
1.0000
0.1666
0.0000
1.0000
0.3333
0.8333
0.3333
0.0000
0.6666
0.6666
0.3333
0.6666
0.3333
0.1666
0.8333
0.8333
0.3333
0.6666
0.0000
0.6666
0.3333

acta computare Volume 1, Year 2012, 242-251

249

Table 1
ABC classifications by different models
Item

ADU

AUC

LT

sI

S4
S1
S3
S6
S7
S5
S8
S11
S16
S25
S24
S17
S41
S26
S15
S30
S42
S32
S21
S12
S36
S46
S44
S35
S9
S20
S38
S22
S19
S37
S23
S47
S43
S33
S39
S18
S31
S14
S10
S40
S28
S34
S13
S27
S45
S2
S29

5840.64
5037.12
4769.56
2936.67
2820
3478.8
2640
1075.2
1043.5
2423.52
703.68
190.89
25.38
463.6
5670
810
79.2
398.4
338.4
370.5
150
570
59.78
28.8
594
34.4
75.4
163.28
854.4
467.6
197.92
212.08
48.3
103.36
224
455
1038
181.8
119.2
134.8
216
313.6
432.5
336.12
2407.5
883.2
268.68

49.92
23.76
27.73
31.24
28.2
57.98
55
5.12 1
20.87
73.44
14.66
7.07
8.46
24.4
10
45
19.8
33.2
33.84
37.05
30
47.5
29.89
28.8
49.5
34.4
37.7
40.82
71.2
58.45
49.48
53.02
48.3
51.68
56
65
86.5
60.6
59.6
67.4
72
78.4
86.5
84.03
160.5
110.4
134.34

2
4
1
3
3
3
4
2
5
6
4
7
5
4
5
3
2
3
3
1
5
5
5
3
6
7
2
3
3
4
5
2
3
6
1
4
7
3
5
3
5
6
4
1
4
5
7

1.0000
0.9924
0.9446
0.7021
0.6984
0.6748
0.5802
0.5615
0.5014
0.4873
0.4809
0.4435
0.4178
0.4160
0.4084
0.3862
0.3809
0.3765
0.3668
0.3593
0.3567
0.3471
0.3460
0.3458
0.3430
0.3397
0.3200
0.3199
0.2984
0.2953
0.2933
0.2826
0.2788
0.2737
0.2733
0.2703
0.2669
0.2517
0.2474
0.2217
0.2154
0.2049
0.1909
0.1876
0.1753
0.1624
0.1523

ABC classification
proposed model
A
A
A
A
A
A
A
A
A
A
B
B
B
B
B
B
B
B
B
B
B
B
B
B
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C

ZF
C
A
A
C
C
B
B
C
C
C
C
C
C
C
C
C
C
C
C
B
C
C
C
C
B
C
B
A
B
B
B
C
C
B
B
A
B
A
A
B
A
B
A
C
B
A
A

H.A
A
A
A
A
B
A
B
C
C
C
C
C
C
C
C
C
C
C
C
B
C
C
C
C
C
C
C
A
B
C
B
C
C
B
B
B
B
B
A
B
B
B
A
C
B
A
A

Ng
A
A
A
A
A
A
A
A
B
B
B
B
B
B
C
C
C
C
B
A
C
B
C
C
C
C
C
A
B
B
C
B
B
C
C
C
C
C
C
C
C
B
C
C
C
B
C

Traditional
A
A
A
A
A
A
A
B
B
C
B
B
C
C
C
C
C
C
B
B
C
C
C
C
B
C
B
A
B
C
B
C
C
C
C
B
C
B
A
C
C
C
B
C
C
A
C

250

acta computare Volume 1, Year 2012, 242-251

4.Conclusion
In this paper, a simple approach for inventory classification is proposed. A weighted
nonlinear programming model has been proposed and illustrated in this paper for
classifying inventory items in the presence of multiple criteria. It is a very simple model
that can be easily understood by inventory managers.
References
[1] C.H. Cheng, Evaluating weapon systems using ranking fuzzy numbers, Fuzzy Sets
and Systems, 107, 25-35, (1999).
http://dx.doi.org/10.1016/S0165-0114(97)00348-5
[2]

B.E. Flores, D.L. Olson and V.K. Dorai, Management of multicriteria inventory
classification, Mathematical and Computer Modelling 16 71-82 (1992).
http://dx.doi.org/10.1016/0895-7177(92)90021-C

[3]

M.A. Cohen, R. Ernst, Multi-item classification and generic inventory stock control
policies, Production and Inventory Management Journal 29 (1988).

[4]

R. Ernst, M.A. Cohen, Operations related groups (ORGs): a clustering procedure for
production/inventory systems, Journal of Operations Management, 9, 574-598,
(1990).
http://dx.doi.org/10.1016/0272-6963(90)90010-B

[5]

T.L. Saaty, The analytic hierarchy process, McGraw-Hill: New York; (1980).

[6]

F.Y. Partovi, J. Burton, Using the analytic hierarchy process for ABC analysis,
International Journal of Production and Operations Management, 13, 29-44,
(1993).
http://dx.doi.org/10.1108/01443579310043619

[7]

F.Y. Partovi and W.E. Hopton, The analytic hierarchy process as applied to two
types of inventory problems, Production and Inventory Management Journal, 35,
13-19, (1993).

[8]

H.A. Guvenir, E. Erel, Multicriteria inventory classification using a genetic

algorithm, European Journal of Operational Research, 105, 29-37, (1998).
http://dx.doi.org/10.1016/S0377-2217(97)00039-8

[9]

F.Y. Partovi, M. Anandarajan, Classifying inventory using an artificial neural

network approach, Computers and Industrial Engineering, 41, 389-404, (2002).
http://dx.doi.org/10.1016/S0360-8352(01)00064-X

[10]

R. Ramanathan, ABC inventory classification with multiple-criteria using weighted

linear optimization, Computers & Operations Research, 33, 695-700, (2006).
http://dx.doi.org/10.1016/j.cor.2004.07.014

acta computare Volume 1, Year 2012, 242-251

251

[11]

P. Zhou, L. Fan, A note on multi-criteria ABC inventory classification using weighted

linear optimization, European Journal of Operational Research, 182, 1488-1491,
(2007).
http://dx.doi.org/10.1016/j.ejor.2006.08.052

[12]

W.L. Ng, A simple classifier for multiple criteria ABC analysis, European Journal of
Operational Research, 177, 344-353, (2007).
http://dx.doi.org/10.1016/j.ejor.2005.11.018

[13]

K. Rishi, R.K. Garg, Optimal selection of robots by using distance based approach
method, Robotics and computer-Integrated Manufacturing, 26, 500-506, (2010).
http://dx.doi.org/10.1016/j.rcim.2010.03.012