Cost-Effective Optimization on a Two Demand Classes Inventory System

Int J Appl Comput Math (06) :57 73 DOI 0007/s4089-05-0046-6 ORIGINAL PAPER Cost-Effective Optimization on a Two Demand Classes Inventory ystem Raj Kumar Malligarjunan Published online: 9 March 05 pringer India Pvt Ltd 05 Abstract This paper investigates impact of change of priority on a continuous review inventory system with a mixture of backorders and lost sales The inter-arrival processes are assumed to be exponentially distributed The objective is to minimize the anticipated total cost rate by simultaneously optimizing the maximum inventory level and backorders An entropy-based uncertainty metric for evaluating the operating performance is demonstrated The proposed modification offers the practical value of establishing how the long-run cost rate is reduced Various system performance measures are derived Keywords tochastic inventory Poisson demands Lost sales Backorders Multiple demand classes Mathematics ubject Classification 90B05 60J7 List of symbols s λ λ L(t) φ i E(L) E(B) E(R) E(R ) Maximum inventory level Maximum backorders Arrival rate of type-(high priority) demands Arrival rate of type-(low priority) demands On-hand inventory level at time t teady state probability at state i Mean of on-hand inventory Mean of backorders Expected reorder order Expected reorder order in priority scheme B Raj Kumar Malligarjunan mrajmkumarm@gmailcom Devanga Arts College, Aruppukottai, Tamilnadu, India 3

58 Int J Appl Comput Math (06) :57 73 E() V (L) V (B) A B m m m L F(t) α h K c B c TC TC E H( ) TC Expected shortage rate Variance of on-hand inventory Variance of backorders tate space of inventory level tate space of backorders Expected time between two reorder times in priority case Expected time between two reorder times in equal-priority case Expected time between two reorder times in non-priority case Distribution of time of reorder process Initial probability vector Holding cost per item per unit time etup cost per order Backorder cost per customer per unit time Losses in revenue per customer per unit time Total expected cost per unit time Total expected cost per unit time in equal priority hannon entropy for Relative cost decrease Introduction A matter that is regarded as more important than another is called as priority In many inventory settings, a supply firm wishes to provide different levels of service to different customers For example, in a service-parts network, a customer can choose amongst different contracts, each with a different price and level of service o, there is a need to study inventory systems with different demand types Arslan et al considered a single-product inventory system that serves multiple demand classes, which differ in their shortage costs or service-level requirements They assumed a critical-level control policy, and a backorder clearing mechanism in which treat a backorder for a lower-priority class equivalent to a reserve-stock shortfall for the high-priority class Tang et al considered a two-class model in which high priority was given to the backorder demand class, and low priority was given to the lost-sales demand class The authors in 3 discuss a two-class system with a supply capacity, in which high priority is assigned to the class with lost-sales, whereas part of the unsatisfied demands are lost in the low priority class and the rest are backlogged for one period They took for granted that the replenishment decision must first fully satisfy all previous backorders at the beginning of each period, with the remaining on-hand inventories used to satisfy new requirements anaand Goyal 4 considered a stochastic economic order quantity (EOQ) model over a finite time horizon where uniform demand over the replenishment period is price dependent The author developed the criterion for the optimal solution for the replenishment size such that the integrated expected profit is maximized An intuitionistic fuzzy EOQ inventory model with backlogging was investigated using the score functions for the member and nonmembership functions by De and ana 5 Pal et al6 analyzed a single-period newsvendor model to determine the optimal order quantity where the customers balking occurs The model allowed partial backlogging and permits part of the backlogged shortages to turn into lost sales arkar and aren 7 considered the strategy that supplier offers retailer a full trade-credit policy whereas retailer offers their customers a partial trade-credit policy The 3

Int J Appl Comput Math (06) :57 73 59 main purpose was to minimize the retailers annual total cost with finite replenishment rate A comparison between inventory followed by shortages model and shortages followed by inventory model with variable demand rate was studied by Khanra et al 8 The writing of Cárdenas-Barrón et al 9 provides a basis for new directions in inventory management research For further related works, our readers may refer 0and Isotupa analysed a two demand classes inventory system with lost sales he assumed that low priority customers are lost when the inventory level is low Zhao and Lian 3 addressed a queueing-inventory system with two classes of customers in which the priority for service is decided by the server Liu et al 4 analysed a flexible service policies for an inventory system with two classes of customers When the on-hand inventory drops to pre-determined safety level, arrival ordinary customers receive service at a fixed probability Raj Kumar 5 considered a two demand classes perishable inventory system that consists of a single server He analysed an entropy-based uncertainty metric for measuring the operating performance Lost sales are those selling opportunities that the shopkeeper has lost because an item was out of stock that caused him to lose the opportunity to sell For example, product wanted not in the assortment, size out of stock, could not find product wanted, product out of stock, and did not receive help are some reasons for lost sales The primary cost control focus has been on enhancing the order processing system not only to eliminate excessive production of inventory but also to decrease the number of stock outs and lost sales Lost sales are a difficult risk to put a monetary value on, but we all must understand the impact and risk o, this work concentrates on lost sales We would like to find what happens when there is no item in the stock and the impact of prioritization during stock out periods Moreover, it is well known that many systems operate subject to random occurrences Entropy is a measure of randomness, flexibility, order/disorder and uncertainty It appears in the laws that are fundamental to several fields of science for example data compression and transmission, ergodic and probability theory For more details about entropy measures, see Cover and Thomas 6 In this paper, we study a continuous review inventory system serving two demand classes, in which low priority is assigned to the class with lost sales, while high priority is assigned to the class with backorders Compared with earlier works, our paper assumes, instead of loosing one type of customers, both types of demand classes are allowed during stock out period with equal priority The results show that the modification is cost-effective The remaining part of this paper is prepared as follows: The Model is described in second section The third section provides the analysis of the inventory level and the number of back orders at steady state In fourth section, the analysis of reorder times is given In fifth section, several related performance measures of significance are provided Based on these measures, we calculate the long-run total expected cost function which is pseudo-convex in both parameters s and and the results are compared with existing models The measures on uncertainty is provided in sixth section ome numerical results are given in the last ection Model We consider a continuous review inventory system The capacity to store inventory is The items are drawn one at a time as and when a demand occurs There are two types of customers namely type- and type- The customers arrive at the system independently according to 3

60 Int J Appl Comput Math (06) :57 73 two Poisson processes with rates λ and λ A shopkeeper is always available in the system who serves to the demands one by one If there is a shortage of s units, an order for + s units is placed and the replenishment of stock is instantaneous When the inventory level is zero, the shopkeeper allows only one half of customers from the two demand classes and the other half are assumed to be lost teady-tate Probability Distribution uppose L(t) denote the on-hand inventory level at time t, φ i (t) = P L(t) = i so that the steady state probability is We have, φ i = lim t φ i (t) φ i (t) = λ φi+ (t) φ i (t), i =,,,, φ i (t) = λ φ i (t) = λ In the steady state φ i+ (t) φ i (t), i = 0, φi+ (t) φ i (t), i =,,, s + { φ0 i =,,, s +, φ i = φ 0 i =,,, Now, multiplying Z i with φ i and summing for all i,weget () and s+ i=0 φ i z i = φ i z i = i= s+ i=0 i= The probability generating function is then defined as (z) = i= s+ = φ 0 i= φ i z i φ 0 z i φ 0 zi z i 0 + i= s+ and the value of φ 0 can be obtained from lim z (z) = as 3 φ 0 = s + z i

Int J Appl Comput Math (06) :57 73 6 To obtain the mean and variance of on-hand inventory level, we define (z) = φ i z i By taking derivatives of the above equation at z =, we obtain the following: ( + ) E(L) = φ 0 4 and V (L) = i(i )φ 0 z i E(L) i= ( + ) 4 ( + ) = φ 0 φ 0 6 8 imilarly, to get the mean and variance of backorders, consider s (z) = φ 0 z i Thus i= i= s(s ) E(B) = φ 0 s(s ) and V (B) = φ 0 3s s + 5 6 We note that for = ands =, the process {L(t); t 0} is a continuous time Markov process with state space = A B = { s +, s +,,0} {,,,} The Chapman Kolmogorov forward equations are then given by φ (t) = λ φ (t) + φ 0(t) φ 0 (t) = λ φ 0 (t) = λ φ 00 (t) = λ φ (t) φ 0 (t) φ 0 (t) + φ 00 (t) φ 0 (t) φ 00 (t), φ (0) =,, φ 0 (0) = 0, We solve the first two forward equations as follows: We have, φ 0 (0) = 0,, φ 00 (0) =, () φ (t) + φ 0 (t) = Pr (L(t) = 0or/L(0) = ) = (3) ubstituting for φ 0 (t) in the first equation, we obtain φ (t) = λ φ (t) + ( φ 00 (t)) λ ( = λ 3 φ (t) + ) (4) 3

6 Int J Appl Comput Math (06) :57 73 This is a non-homogeneous first order differential equation with constant coefficients By solving we obtain φ (t) = ( ) + e 3 λ t (5) 3 Using the Eq (3), we obtain φ 0 (t) = φ (t) = ( ) e 3 λ t (6) 3 imilarly, we can solve the last two equations to obtain φ 0 (t) and φ 00 (t) Combining all these equations, the matrix (t) is given by (t) = 3 ( ) + e 3 λ t 3 ( ) e 3 λ t ( ) e 3 λ t 3 ( ) + e 3 λ (7) t 3 Analysis of Reorder Times In this section we consider the time of reorders As the inventory level depletes from s + to s, the shopkeeper places an order for + s items which occurs when type- or of type- customer demand items and the number of backorders is s We define m as the expected time between two successive reorders in equal priority case uppose m as the expected time between two successive reorders in priority case (see 5) Thus m = m + (on-hand inventory) + m (backorders) = + β i i β A i=i i B = λ + s λ The expected time between two successive reorders is then reduced from the( priority case ) while we apply the equal priority policy The reduced quantity is obtained as s λ λ If m L denotes the expected time between two successive reorders in normal case, then it is the lower bound and we obtain m > m > m L = (s + ) λ uppose E(R) denotes the expected reorder rate and is given by E(R) = λ s+ It agrees the result of 5whenμ = 0 If E(R ) is the expected reorder rate under priority scheme, then λ s + λ λ < λ s + λ while λ >λ and hence we obtain E(R) <E(R ) We also note that suppose 0 = ζ 0 <ζ <ζ < be the times at which reorders are placed ince the supply of orders is instantaneous, the on-hand inventory level at time ζ n (n = 0,,)is and {ζ n, n = 0,,} is a renewal process Then the process obtained is a s + + states continuous time Markov chain in which the state s is an instantaneous 3

Int J Appl Comput Math (06) :57 73 63 state The process is obtained by requiring the path functions to be right continuous with states { s +, s +,, } and F(t) be the distribution function between reorder times As such F(t) is same as the distribution function of the time till absorption into the state s, starting from the initial state Hence, F(t) is the distribution of phase type and {ζ n } is a renewal process of phase type (PH-renewal processes) The nonnegative random variable X of reorder times has a phase-type distribution (PHdistribution) and its distribution function is then given by F(t) = Pr{X t} = αexp(tt)e, t 0 where α is a substochastic vector of order + s, ie,α is a row vector, all elements of α are nonnegative, and αe, and T is a subgenerator of order + s, ie, T is a square matrix of order + s matrix such that T = 0 s + s + λ λ 0 0 0 0 0 0 0 λ λ 0 0 0 0 0 0 0 0 0 λ λ 0 0 λ 0 0 0 0 0 0 0 0 s + λ 0 0 0 0 0 0 λ s + 0 0 0 0 0 0 0 The expected time, m between two successive reorders is then given by m = αt e (see 5) We assume that { λ ; i β i = λ ; i 0 If the process starts at the state, weobtainα = (, 0, 0,,0), and λ m = i = + s (8) β i λ λ uppose the process starts at any one of the states in, the initial probability vector is α = +s et +s,then m = ( + i ) n β i i = + + s β i i i = i β i ( + s)λ ( + ) + 4(s) + s(s + ) (9) 3

64 Int J Appl Comput Math (06) :57 73 Once the shopkeeper starts the system with positive inventory level, the value of α is then given by ( e T, 0, 0,,0), and hence m = + i + β i= i β i i B = + + s (0) λ ystem Performance Measures In this section, we derive several key performance measures of the considered system We construct the long-run total expected cost per unit time using these measures Let E(L) denote the expected inventory level at steady state As φ i denotes the steady state probability of the number items in the system, the expected inventory level is given by E(L) = ( + ) s + 4 Let E(B) denote the expected number of backorders in the steady state E(B) = i= s+ iφ i s(s ) = φ 0 3 uppose E(R) denotes the expected reorder rate, the quantity of it is then given by E(R) = λ s + 4 We observe that the process stays at several states sometimes The proportion of time that the on-hand inventory level is positive is then given by i A i βi = β i λ λ + s λ / = + s, and the proportion of time that the on hand inventory level is negative (backorders) is then given by βi i B = β i i s + s It is noted that the proportion quantities are independent of arrival rates 5 Let E() denote the expected shortage of the customers which is given by 3 E() = sλ s +

Int J Appl Comput Math (06) :57 73 65 To study the cost structure of the model, the above system performance measures are used For this, first we consider the following costs h: the holding cost per unit item per unit time K : set-up cost per order c B : the backorder cost per customer per unit time c : the losses in revenue per customer per unit time The long-run total expected cost per unit time for this system at steady state is then obtained as follows: TC(s, ) = he(l) + KE(R) + c B E(B) + c E() ( + ) λ s(s ) sλ = h ( ) + K + c s + B φ 0 + c 4 s + s + h( + ) = ( ) + K λ s + 4 + c sλ s(s ) + c B Now, we must determine the optimal values of s and to minimise the long-run expected cost per unit time The following two theorems help us to detect the convex nature of TC Theorem For a fixed value of backorders, the long-run expected cost rate is pseudo-convex in Proof Consider T () = TC(s, ) for any fixed s The cost function is then defined as follows: T () = h (+) + K s + with K =K + sc λ + c B s(s ) The first difference = α() β() (say) T () of T () is T ( + ) T () and can be expressed as α () β () where α () = α( + )β() α()β( + ) and β () = β( + )β() As β () is an increasing function of, it is enough to show α () is an increasing function of Thevalue s of α () is a difference of K + h(+)(+) 4 + s and K + h(+) 4 + + Its value is h(+) 4 s + K Itisanincreasingfunctionin This fact completes the proof Theorem For a fixed value of, the long-run expected cost rate is pseudo-convex in s Proof The proof is simple as in Theorem Comparison In this subsection, we derive the total cost function of the inventory system whereas higher priority is given to the back-order demand class and lower priority is given to the lost-sales demand class in two demand classes models (see ) We then compare it with our current model As λ λ with the value of φ 0 = s + λ λ, the waiting time cost rate is given by ( ) s + λ λ c sλ + c B s(s ) () 3

66 Int J Appl Comput Math (06) :57 73 and when λ λ, the other cost rate is h( + ) ( ) s + λ λ ( λ λ ) + K λ () We now define the total cost rate of current model with equal priority by TC E h( + ) = ( ) + (K + sc s + ) λ 4 + c s(s ) B (3) and by utilizing Eqs ()and(), the total cost rate of existing models is given by ( ) h( + ) λ s(s ) TC = ( ) + K λ + c sλ + c B (4) s + λ λ λ Now, we state a lemma to compare the potential values of expected total waiting costs per unit time It is noted that the intensity of arrival rate of type- is significant on the total cost rate cb s(s ) 4λ Lemma Assumed that ζ = s c s + c and ζ = Ifthearrivalrateof type- customers is higher than the arrival rate of type- customers, then ζ,ζ > 0 and the expected waiting cost of T C E is smaller than of T C Proof 3 We prove this lemma by showing the expected waiting costs of TC TC E > 0 Consider the difference of expected waiting costs of these two functions as s + (sc λ ) s + λ ( ) λ sc λ Thus s c s + c (λ λ )>0 (5) imilarly, the difference of the backorder costs is c B s(s ) ( ) ( ) s + λ s + λ which is given by cb s(s ) 4λ (λ λ )>0 (6) The assumption implies that the Eqs (5) and (6) are true and this completes the proof We now ready to state our main result comparing the waiting costs in the total cost rate Theorem 3 Assume that ζ 3 = sk and ζ 4 = sh(+)λ 4 and ρ = ζ 3+ζ 4 ζ +ζ If the arrival rates satisfy the condition of Lemma,theTC E is cost-efficient than the other values of T C 3

Int J Appl Comput Math (06) :57 73 67 Proof 4 As in the above Lemma, the comparison of the set-up cost rates yields ζ 3 from { s + K λ s + λ K λ } (λ λ ) λ and the holding cost rates give ζ 4 from the expression { s + h( + )λ s + λ } h( + ) (λ λ ) λ λ 4 On collecting the values of ζ s, as ρ<, the expected total cost rate of TC E < TC when the condition of the Lemma holds, we obtain the result Measure on Uncertainty The concept of the steady-state of an inventory/queue concerns a probability function, it seems logical to consider a connection between entropy and the uncertainty in this model Indeed, hannon entropy 7 considers the probability distribution of signals transmitted over a given communication channel in its argument of uncertainty Therefore, it is found feasible and logical that the entropy measure would give the uncertainty of the model behaviour in the long-run The hannon entropy for the probability vector of this mode is defined as We now define H( ) = i= s+ φ i log φ i = sφ 0 log φ 0 ( = φ 0 log = φ 0 log φ 0 ( φ ) 0 log φ 0 ( ) ) s + log φ 0 { ( ) ( ) s+ } φ 0 h(φ i ) = φ i ; (0 φ i ) (7) We note that if φ i, i =,,,, thenh(φ i ) in (7) will be relatively large and if φ i, i = s +, s +,,0, then h(φ i ) will be small Moreover, the principle of maximum entropy suggests what to do and how should the assignment of probabilities be changed when we happen to know (or learn) something about the non-uniformity of the arrivals To maximize the entropy among all probability distribution on the set, we want to optimize (maximize) H( ) = φ i 0and φ i = We assume i s+ φ i log φ i with the constraint 3

68 Int J Appl Comput Math (06) :57 73 Table ummative result of cost and entropy rates Cost rates h K c B c Entropy rate ( λ Priority (P) (+) s(s ) λ λ sλ log s + λ )( ) λ (s+λ /λ ) λ λ ( (+) λ Equal priority (EP) s(s ) 4 s λ ) log s + (s+/) Normal (N) (+) λ s(s ) log(s + ) Ĥ( ) = H(, ξ) ( ) = φ i log + ξ φ i, (8) φ i i= s+ i where ξ is a Lagrange multiplier and the φ i s are probabilities ince Ĥ φj = + ξ + log φ j, so φ j = e ξ for all j It happens on two cases Case For λ λ, i A and λ = 0, i B, we obtain the maximum entropy distribution ˆ which is a discrete uniform distribution of size s + inceλ >λ,the arrival rate of type- customer is reduced Case When a quantity of λ is added with the arrival rate for i A, we also obtain the same maximum entropy distribution ˆ By the results of the cases, we note that H( ) = φ 0 log { ( ) ( ) s+ } φ 0 log(s + ) (9) To compare these results, we tabulate the cost rates and the corresponding entropy rates as follows (Table ) The priority case (P) in the table shows the results of cost rates and the corresponding entropy rate The equal priority (EP) case can be directly obtained from P as λ λ The entropy rate of EP is more than the entropy rate of P imilarly, the normal (N) case can be also obtained from P as λ λ The N has the maximum entropy rate of quantity log(s + ) Also, we note that the holding and the set-up cost rates are increased while the case P changes to EP and then changes to N On the other hand, the shortage rate is decreased during the stock-out period This fact can intuitively be explained in terms of lost sales, because the shortage of non-priority customers is reduced We now consider the steady state distribution with non-perishable items which was derivedin5 The probability distribution is used to derive the relative entropy It is a measure of distance between the distributions and and is given by 3

Int J Appl Comput Math (06) :57 73 69 Fig The optimal value of D( ) = φ i log φ i φ i i ( ) = sφ 0 log φ 0 φ 0 + φ φ0 0 log ( φ ) 0 β 0 i B β i ( ) s + λ λ = sφ 0 log s + + φ 0 log = sφ 0 log s + λ + φ 0 λ log ( log s + ) (s + )φ 0 + ( s + λ s + ( λ λ ) λ λ λ s + )( λ λ ) Numerical Examples In this section, we report our computational experiments on the behaviour of the optimal cost and the optimal solution Example In the first example, take the values of parameters and costs as λ = 0; λ = 50; K = 300; h = ; c B = 3; c = 0 As in Theorem, wefixthevalueofs in order to find the optimal value of By varying the values of, the minimum amount of total cost per unit time TC E is found and its value is 705 for (5, 70) Figure represents the long run expected total cost rate It contains three curves together corresponding to s =5,6 and 7 We note that the cost is lesser under the higher backorder facility The optimal policy of TC E for s is given in Fig It represents the long run expected total cost rate for s with the same parameter and cost values of the first example It contains three convex curves together corresponding to =0, and As in the first example, we 3

70 Int J Appl Comput Math (06) :57 73 Fig The optimal value of s Fig 3 Maximum entropy for various s fix the value of and vary the values of s The minimum of total cost per unit time TC E is 8649 at (3, ) The system with higher stocking facility obtains the minimum cost Example We now focus our attention to the maximum entropy obtained in ystem performance measures section We fix λ = 0 and λ = 40 and take the various maximum stocking level and backorders We have experimented on the entropy rates and compared the maximum entropies for priority, equal priority and normal cases Figure 3 shows the maximum entropy for the mentioned three cases The minimum value 396059 is obtained at (s = 0, = 00) for priority type and the highest of all the cases is (=500635) obtained for normal type This says the system which attains the higher entropy quantity is more profitable (Fig 4) Example 3 In order to evaluate the performance of this change, we compare the expected long run waiting cost rate of TC E with the existing cost structure in the third example It 3

Int J Appl Comput Math (06) :57 73 7 Fig 4 Influence of on maximum entropy Table The relative cost decrease for lower values of s s 5 6 TC E TC %of decrease TC E TC %of decrease λ = 0 (lower) 40 87889 958 55 9384 07635 548 4 87565 9553 553 93480 06936 5483 4 8743 9449 554 9339 064 5484 43 8693 9383 556 9800 0555 5485 44 86606 9379 557 9464 04867 5487 45 869 959 558 930 0486 5488 46 85978 9884 559 9799 0350 5489 47 85668 944 550 9470 0838 549 48 85360 90607 55 943 07 549 49 85054 89975 553 9089 0508 5493 50 84750 89347 554 90496 00850 5494 follows from the proofs 3 and 4 that TC E < TC The main purpose of the experiment is to study the magnitude of this difference as the priority changes, and explore the conditions under which the cost savings from using TC E instead of TC being most significant We denote then by TC = TC TCE TC, the relative cost decrease when using the modified policy instead of the existing policies It is noted that when increases, the values of TC and TC E decrease and increase as s increases If λ increases, then both the values come together (Tables, 3, 4) The modification seems to perform in all the situations when λ has lower values The tables show that the difference of arrival parameters work rationing phenomena considerably In particular whenever λ is extremely low, the relative cost decrease is more satisfiable It may be noted that these situations are most appropriate for rationing, and that the relative cost decrease will be high 3

7 Int J Appl Comput Math (06) :57 73 Table 3 The relative cost decrease for moderate values of s s 5 6 TC E TC %of decrease TC E TC %of decrease λ = 35 (medium) 40 04556 43775 78 47 593 7 4 0470 4357 78 06 5384 7 4 03787 474 79 0657 584 7 43 03407 43 730 055 5303 73 44 0309 474 730 09855 50768 74 45 0655 4 73 09458 5037 74 46 083 407 73 09065 49709 75 47 093 404 73 08674 4985 76 48 0547 3973 733 0886 48665 76 49 083 394 733 07900 4849 77 50 008 38756 734 0758 47636 77 Table 4 The relative cost decrease for higher values of s s 5 6 TC E TC %of decrease TC E TC %of decrease λ = 35 (higher) 40 NF 98 98 NF 4 0775 0775 NF 8645 8645 NF 4 033 033 NF 875 875 NF 43 9890 9890 NF 7709 7709 NF 44 9453 9453 NF 746 746 NF 45 908 908 NF 6787 6787 NF 46 8587 8587 NF 633 633 NF 47 859 859 NF 5878 5878 NF 48 7734 7734 NF 549 549 NF 49 73 73 NF 498 498 NF 50 6893 6893 NF 4539 4539 NF NF Not feasible Conclusion, Observations and Future directions In this paper, a two demand classes inventory system was considered with modified service policy The steady-state and transient solutions were obtained analytically The expressions for the distribution function of the reorder times were obtained everal system performance measures were derived and the modified cost structure was analysed An entropy-based measures were also calculated We obtain the following managerial suggestions in this analysis: The modification of the existing models suggests that the total cost per unit time is reduced when the arrival rate of type- (ordinary) customers is greater than the arrival rate of type- 3

Int J Appl Comput Math (06) :57 73 73 (priority) The reduction of reordering times suggests better service levels to all types of customers Entropy maximization is equivalent to profit maximization It would be interesting to analyze the problem discussed in this article where demands due to one type of customer are not a Markov process Naturally, with the inclusion of non- Poisson demands, the problem will be more challenging The model can also be extended to include bulk demand for anyone of two types of demands We are now working on the above mentioned extensions, and these will be reported in future publications Acknowledgments The authors are very thankful to all the three reviewers for their valuable suggestions References Arslan, H, Graves, C, Roemer, T: A single-product inventory model for multiple demand classes Manag ci 53(9), 486 500 (007) Tang, Y, Xu, D, Zhou, W: Inventory rationing in a capacitated system with backorders and lost sales In: Proceedings of the 007 IEEE-IEEM, 579 58 (007) 3 Duran,, Liu, T, imchi-levi, D, wann, JL: Policies utilizing tactical inventory for servicedifferentiated customers Oper Res Lett 36, 59 64 (008) 4 ana, : The stochastic EOQ model with random sales price Appl Math Comput 8(), 39 48 (0) 5 De, K, ana, : Backlogging EOQ model for promotional effort and selling price sensitive demandan intuitionistic fuzzy approach Ann Oper Res 03, 0 (03) 6 Pal, B, ana,, Chaudhuri, K: A distribution-free newsvendor problem with nonlinear holding cost Int J yst ci 46(7), 69 77 (05) 7 arkar, B, aren, : Partial trade-credit policy of retailer with exponentially deteriorating items Int J Appl Comput Math (04) doi:0007/s4089-04-009-8 Khanra,, Mandal, B, arkar, B: A comparative study between inventory followed by shortages and shortages followed by inventory under trade-credit policy Int J Appl Comput Math (05) doi:0 007/s4089-05-004-z 9 Cárdenas-Barrón, LE, Chung, KJ, Treviño-Garza, G: Celebrating a century of the economic order quantity model in honor of Ford Whitman Harris Int J Prod Econ 55, 7 (04) 0 ana, : Price sensitive demand with random sales price-a newsboy problem Int J yst ci 43(3), 49 498 (0) ana,, Goyal, K: (Q, r, L) model for stochastic demand with lead-time dependent partial backlogging Ann Oper Res 04, (04) Isotupa, KP: An (s, Q) Markovian inventory system with lost sales and two demand classes Math Comput Model 43, 687 694 (006) 3 Zhao, N, Lian, Z: A queueing-inventory system with two classes of customers Int J Prod Econ 9, 5 3 (0) 4 Liu, M, Feng, M, Wong, CY: Flexible service policies for a Markov inventory system with two demand classes Int J Prod Econ 46, 566 574 (03) 5 Raj Kumar, M: Transient behaviour and entropy measures on an inventory system with two demand classes Appl Math Comput 6, 738 753 (04) 6 Cover, TM, Thomas, Joy A: Elements of Information Theory, nd edn Wiley, Hoboken (006) 7 hannon, CE: A mathematical theory of communication Bell yst Tech J 7, 379 43 (948) 3