A Study on the Chain Restaurants Dynamic Negotiation Games of the Optimization of Joint Procurement of Food Materials

Transcription

1 International Journal of Coputer Science & Inforation Technology (IJCSIT) Vol 6, No 1, February 2014 A Study on the Chain estaurants Dynaic Negotiation aes of the Optiization of Joint Procureent of Food Materials Hung-Teng Chang Departent of Inforation Manageent, Yu Da University of Science and Technology, Taiwan,.O.C. ABSTACT In the era of eager profit, production costs often becoe an iportant factor affecting SMEs operating conditions, and how to effectively reduce production costs has becoe an issue of in-depth consideration for the business owners. Especially, the food and beverage (F&B) industry cannot accurately predict the deand. It any cause deand forecast fall and excess or insufficient inventory pressure. Copanies of the F&B industry ay be even unable to eet iediate custoer needs. They are faced great challenges in quick response and inventory pressure. This study carried out the product inventory odel analysis of the ost recent year s sales data of the fresh food aterials for chain restaurants in a supply chain region with raw aterial suppliers and deanders. Moreover, this study adopted the ulti-agent dynaic strategy gae to establish the joint procureent decision odel negotiation algorith for analysis and verification by siulation cases to achieve the design of dynaic negotiation optiization echanis for the joint procureent of food aterials. Coupled with supply chain anageent 3C theory for food aterial inventory anageent, we developed the optiization ethod for deterining the order quantities of the chain restaurants. For product deand forecast, we applied the coonality odel, production and delivery capacity odel, and the odel of consuption and replenishent based on arket deand changes in categorization and developent. Moreover, with the existence of dependencies between product deands as the deand forecast basis, we deterined the appropriate inventory odel accordingly. KEYWODS Multi-agent, Joint procureent, ae theory, 3C theory 1. INTODUCTION As it is difficult to accurately predict the deand, higher deand uncertainty will be accopanied with increasing costs of forecast errors of the firs. The reason for the difficulties in forecast of the future is that the arket is characterized by uncertainty, coplexity, and copetition. These three characteristics ake any firs unable to grasp the arket deand conditions, resulting in too any inventories to cut sales prices or insufficient inventories to fail to grasp the favorable opportunity to ake oney. The three characteristics of the arket are fair to each fir on the arket. Any fir that can ake the three characteristics ore transparent can have ore business opportunities to reduce losses. This study applied the gae theory in ulti-agent joint procureent decision odel negotiation algorith to find out the relatively better order quantity of a certain raw aterial for chain restraints participating in the joint procureent and achieve the goal of lowering overall operating costs by way of negotiation. We further introduced the concept of 3C theory to use the DOI : /ijcsit

2 International Journal of Coputer Science & Inforation Technology (IJCSIT) Vol 6, No 1, February 2014 coonality odel, production and delivery capacity odel, and the odel of consuption and replenishent based on arket deand changes for categorization and developent of arket deand forecast to help copanies iprove forecast accuracy, and reduce their guesswork cost. 2. LITEATUE EVIEW egarding the MAS negotiation echaniss, the ost influential one is Contract Net Protocol, which is derived fro outsourcing tendering procedures of copanies and organizations. Proposed by Sith [1] in 1980, the protocol is to solve probles and conflicts through ulti-agent cooperation and negotiation ode by siulating the tendering procedures of enterprises. According to the echanis of Contract Net Protocol, through the publication of contract inforation, the ulti-agent integrates resources and behaviors for the coon goal. Issues including the sub-task assignent of agent playing the role of the anager, inforation counication between various agents, the bid evaluation odel of participating agents and the foration of copetition and cooperation odels are iportant topics of Contract Net Protocol. egarding the application of Contract Net Protocol, Cantaessa [2] analyzed the agent behavior odel by distributed artificial intelligence and distributed cooperation, and used the Contract Net Protocol to analyze the ode; in this odel, if an agent has a deand, other agents that can satisfy the deand will copete by bidding. Finally, the agent sending out the deand will evaluate the bid according to tie, cost and other factors to forulate the final decisions. The design of the agent bid evaluation odel is the ajor research topic. Through the design of relevant variables, the agent decision-aking odel can generate better syste perforance and respond to the elasticity of real situation [3]. According to the findings of David [4], no copany can satisfy the deand on resources for production fro within itself and no copany can resist the environental pressure by its own force. Therefore, labor of division will coe naturally in between copanies based on specialty, and copanies will depend on each other, develop with each other to for a counity of gains and losses. In related studies, joint procureent has been widely recognized as a feasible ethod to effectively reduce production costs. Operating entities of the sae nature can achieve the goal of cutting price by increasing order quantity through joint procureent. The eaning of joint procureent is to suarize the deands of copanies in the sae industry or affiliated copanies to place orders with the supplier to collectively change the original arket structure and win the influencing power to change arket price echanis, so that the goal of cutting procureent price by increasing order quantity can be achieved. [5] However, in joint procureent, soe probles that participants of joint procureent ay not be able to address effectively, such as how to deterine the joint purchase cycle, whether the negotiated and agreed batches and order quantities in the purchase cycle can achieve the axiized and relatively high profits. Kaspi and osenblatt [6-7] proposed two study reports of the testing of different heuristic algorith, suggesting the optial solution obtained by using AND procedure proposed by Kaspi and osenblatt is better than the solutions found by using algoriths proposed by other scholars (for details, see oyal [8], Brown [9], oyal and Belton [10], and Silver [11]). Unlike the ethods proposed by the aforeentioned scholars, we applied the gae theory in ulti-agent joint procureent decision odel negotiation algorith to find out the relatively better order quantity of a certain raw aterial for chain restaurants participating in joint procureent. In the study of the behavior of the decision akers, the gae theory is different fro econoics and behavior sciences. When the decision-aking behaviors of the decision akers have direct, 100

3 International Journal of Coputer Science & Inforation Technology (IJCSIT) Vol 6, No 1, February 2014 utual ipact, the choice of one party will be affected by the choices of other parties while other parties will be affected by the choice as well. This is the basic feature of applying gae theory in decision aking and equilibriu probles [12]. The gae theory has the following two basic assuptions: 2.1. ationality assuption If a decision-aker can consistently ake decisions in the pursuit of his objectives, then the decision aker is regarded as rational. The objective of rational decision-aker is to pursue the axiization of his or her individual payoffs. Soe very weak assuptions regarding how rational decision akers should behave can prove that there is a certain way to grant utility values to various possible results of his concern to allow his choice to axiize his own utility [13-14] Intelligence assuption If either play of the gae knows what other players know about the gae and ake inferences of other players regarding the situation, the player is called as intelligent. When rational and intelligent decision akers affect each other, their probles will be analyzed together like an equation. Such an analysis is the just the ajority of the gae theory [13-14]. The coplexity of decision-aking probles, and the speed, reliability, flexibility, openness and re-configurability requireents of addressing such probles of the supply chain anageent ake it ipossible to solve with a single agent. It requires the coordination of ultiple agents with knowledge and tasks of different fields to support the supply chain decision aking process. MAS (Multi-Agent Syste) consists of ultiple agents to realize the overall function or purpose by the collective activities consisting of the proble-solving activities of the agents and the interactive activities of agents. Meanwhile, each independent agent realizes its own function and purpose during the interactive process [15]. Therefore, for the coordinated, distributed and integrated requireents of the decision-aking probles of the supply chain anageent, MAS technology can be applied in study. MAS syste uses the distributed and cooperative agents to solve the probles of various nodes. It has its own purposed, knowledge and ability. Agent can be huan, software or intelligent syste. The agents of the MAS syste can coordinate their behaviors to jointly solve the coplex probles [16]. In the working environent of MAS, a coordination agent can be designed to coordinate the work in between groups and reduce the conflicts. Through the inforation exchange of ultiple agents, collective negotiation and decision aking can be realized to get the ost satisfactory answers of the probles. Therefore, MAS syste provides an effective way for the research and developent of coplex supply chain decision aking support systes. The collaborative decision-aking syste of supply chain is presented by the hierarchical decoposition and sharing of decision aking activities. Based on joint orientation, ultiple agents coitted to sharing the coputation activities for the federation defined as the following equation with four unknown quantities: Agent_federation (anager, intra_agent, extra_agent, constructor) The above equation includes a anager (as the federation anageent agent), a few infra_agents ( 0 ) and extra_agents ( >0) with corresponding counication support functions. The anager adinisters the coordination of the federation in a centralized way while no counication is ade in between infra_agents and in between extra_agents. The infra_agents belong to the anager, which can directly assign tasks to infra_agents. The extra_agents belong 101

4 International Journal of Coputer Science & Inforation Technology (IJCSIT) Vol 6, No 1, February 2014 to other anagers and they are teporary ebers of the federation. The anager can assign tasks to extra_agents after consultation with the [17-18]. Developed by Lucent Technical Corporation, 3C theory is the basic theory for the realization of global supply chain anageent and the basic theory for the planning and realization of global resource plan. The basic concepts and values of coonality odel is to achieve the goals of reducing developent costs, siplifying resource anageents, lowering inventories and providing ore diversified products to custoers through expanding the use of coon aterials or resources strategy and product cobination planning. In expanding the use of coon aterials or resources, there are two ajor aspects, naely, odular design and the design for assebly and aintenance. The basic concepts and values of the production and delivery capacity odel is that production capacity, supply capacity or delivery and transportation capacity of copanies of the supply chain have constraints by applying the TOC (Theory of Constrain), and the adjustent and distribution of all resources are ade with such constraints. This is basic spirit of TOC. The production and delivery capacity odel is characterized by the upper liits of aterials and production capacity deand. [19] Whether aterial and production capacity can eet the coitent to the custoer should be taken into consideration when placing the order. Hence, it is a realistic odel to reduce delivery delays caused by lack of inventory or lack of production capacity to enhance custoer satisfaction. The functions designed according to the production and delivery capacity theory, we can lower the risk generated by aterial shortage and excessive coitent to custoer orders. The basic concepts and values of the odel of consuption and replenishent based on arket deand changes are to realize the echanis to purchase aterials before being deanded through real tie arket inforation cobined with arket deand replenishent odel, and thereby lowering inventory levels, reducing fund reserves and discount losses of inventory products. [20] The core of the odel of consuption and replenishent based on arket deand changes consuption is the arket deand forecast odel because one of the causes of inventory is wrong procureent behavior. The reason of wrong procureent is usually caused by istaken assessent of arket deand. The consuption aterial replenishent odel is characterized as the aterial planned production is based on the real deand changes of the arket with a focus on consistency with arket deand to lower inventory to iniize business operational risks [21]. 3. ESEACH METHOD In this study, in a supply chain region containing the raw aterial suppliers and deanders, we applied the 3C theory in collaborative replenishent anageent, and the gae theory in the joint procureent echanis to establish a ulti-agent joint procureent negotiation odel for analysis and verification by using the siulation case to provide a reference for the selection of business negation strategy. The research architecture is as shown in Figure 1, the research ethods and steps are elaborated as follows: 102

5 International Journal of Coputer Science & Inforation Technology (IJCSIT) Vol 6, No 1, February 2014 Figure 1. Chain estaurants Food Materials Deand Forecasting and Joint Procureent Negotiation MAS Fraework Step 1: According to the 3C theory, to establish the 2: purchase quantity algorith for food aterial inventory anageent with profit axiization as the goal; to establish the forecasting and estiation syste of raw aterial deand for various restaurants based on the data of previous order quantities. By referring to the order strategy, the recoended procedure of order policy option odel is as shown in Figure 2, the steps of the procedure are elaborated as follows: Step 1:Materials used in Step Proceed Step3: Select the ABC analyze of corresponding storage the independence and aterials value and odel by aterials dependency needs analysis usage deand ratio category Figure 2. Order Policy Option Model Step 1-1: Part use independence and dependency deand analysis 1.To decide the aterial deand forecasting in the inventory syste 2.Independent deand: deands on different ites are unrelated to each other 3.Dependent deand: the deand on the ite coes fro the deands of other ites Step 1-2: To conduct ABC analysis of the value and use deand rate of various parts. 103

6 International Journal of Coputer Science & Inforation Technology (IJCSIT) Vol 6, No 1, February 2014 When resources are liited, we need to ake the ost effective use of liited resources to control inventory. According to Pareto theore, we should concentrate on the ost iportant inventory ites. It is ipossible to provide a odel or fully control each product. ABC categorization ay be carried out first. Step 1-3: To select the corresponding inventory odels according to the food aterial categorization. As fresh food aterials can easily decay, coupled with shelf life considerations as well as the differences in purchase and acquisition, the inventory odels should be selected according to specific characteristics. It is assued the lead tie of aterial fro the supplier to the factory is LT. 1. For A category product or B category product of COMI >= 0.5, S-s order control syste is applied as shown in Figure 4: (1). Tie between purchase can be acquired by using the following equation: TBP = EOQ / BOM (2). Order-Up-to-Level : S = OUT = BOM TBP + SS. (3). Safety inventory (s) = SS = average daily sales volue LT + Z α σ ( α = 0.05) (4). Material s order point: OP = BOM LT + SS. (5). Once the order tie coes, order by referring to the then inventory and the subtraction of EOQ (Econoic Order Quantity) by the then inventory. OUT EOQ OP SS TBP Figure 3. egular Order Model 2. For B category product with COMI < 0.5 or C category product, regular quantitative order odel is applied: when the inventory drops to the order point, the order of EOQ will be placed. 2BOM K (1). EOQ = H (2). eorder Point: OP = BOM LT + SS. 3. For B category product with COMI < 0.5 or C category product, S-s order control syste is applied. 104

7 International Journal of Coputer Science & Inforation Technology (IJCSIT) Vol 6, No 1, February 2014 (1). Maxiu inventory (S) = BOM TBP. SS (2). Safety inventory (s) = = average daily sales volue (3). e-order point: OP = BOM LT + SS. LT + Z σ ( α = 0.05) α Step 2: Use dynaic strategy gae to establish the joint procureent decision odel negotiation algorith for joint procureent fro raw aterial suppliers. The ethod is as elaborated as follows: Step 2-1: Forulate negotiation strategy The satisfaction and expectation of each chain restaurant is the ajor factor for the success of bargaining negotiations. Therefore, only by understanding the preferences and expectations of players, can we participate in bargaining negotiations in the entality of players. It sees that different chain restaurants have subordinate relationships; they have to cooperate to get greater profits. Next, we developed the analysis fraework of negotiation strategy according to past negotiations between various chain restaurants as shown in Figure 4. Figure 4.Negotiation Strategy Fraework After each chain restaurant has conducted strategic negotiations with each other, we suarized by the proposed coordination agent odel to ensure each chain restaurant to achieve the final cooperation goal, expecting to get the optial strategy cobination of axiu copensation. Step 2-2: To apply gae theory in joint procureent decision odel negotiation algorith. 1. Chain restaurant inventory odel, increental order revenue odel and the odel of applying negotiation decision gaes in increental order revenue odel are illustrated as shown below: (1). ( Q, s, S jit, ) Inventory odel is applied as the chain restaurant inventory odel, where 105

8 International Journal of Coputer Science & Inforation Technology (IJCSIT) Vol 6, No 1, February 2014 A. Q is the average forecast deand of chain restaurant i on product j at tie t in the t following inspection cycle, + ~ Q = D t= t+ 1 B. s is the safety inventory of product j of chain restaurant i at tie t at 95% service level, ( σ σ σ + ) s = z0.025 ρ = C. S is the axiu inventory level of product j of chain restaurant i at tie t at 97% service level, S = Q + 2 ρ D. is inspection cycle (2) Increental order revenue odel of chain restaurant i at tie t P F T S P NPV = P (1 + h ) C ) Q C C C ax{0, Q IP Q }, where, ( ij ij ij ij ij 1 A. h ij is the Holding Cost ate of product j for chain restaurant i B. C is the unit purchase cost of product j for chain restaurant i C. C, C, C is the fixed cost, transportation cost and shortage cost of product j for chain F ij T ij S ij restaurant i IP is the initial inventory level of product j for chain restaurant i at tie t D. 1 E. P Q is the order quantity of product j for chain restaurant i at tie t, where, Q + s Q = ax{ Q + s IP 1,0} S P (3) Apply the negotiation decision aking gae in increental order revenue odel NPV = P Q ij P new new P E[ P( Q > Q )] C (1 + h ) Q + ( C C )( Q + Q ) A. Q is the increental order quantity of product j for chain restaurant i in the negotiation decision aking gae at tie t B. new C is the new order cost of product j for chain restaurant i at tie t P C. E [ P( Q > )] is the noral distribution odel s expected value of Q > probability Q P Q of product j for chain restaurant i in the negotiation decision aking gae at tie t, P P where, E [ P( Q > Q )] = Q N( Q > Q ), the noral distribution odel of chain ~ 2 2 restaurant i is N( Q D, σ = ( σ σ ) /( 1) ) = t= Negotiation gae strategic rules (Supply > Deand): (1)To calculate the increental order quantity Q jt 106

9 International Journal of Coputer Science & Inforation Technology (IJCSIT) Vol 6, No 1, February 2014 (2)To send the increental order inforation to various chain restaurants (3)For chain restaurants, according to the extra-procureent revenue odel, NPV the tolerable axiu increental order quantity. Q, to calculate (1)To establish the list of players of the negotiation gae (2)To establish the cooperative federation copensation atrix of each player according to the player list as follows: To deterine the doinator by order and the rest ebers are allies. A. If Q Q jt, to calculate the copensation of the doinator i NPV ( Q ), and the reaining increental or quantity otherwise, suppose Q = Q jt Q jt = Q jt Q, ove to Step (C);, to calculate the copensation of the doinator i NPV Q ), then the return atrix of the doinator i against allies of the cooperative ( federation is (0,..., NPV,0,...,0), ove to Step (E). B. To select the doinator k fro the cooperative federation into the negotiation gae, if Q, to calculate the copensation of the doinator k NPV Q ), and to Q jt calculate the reaining increental order quantity (D); otherwise, suppose Q = Q jt Q jt = Q jt Q (, ove to Step, to calculate the copensation of the doinator k NPV Q ), and the add the copensation atrix of the doinator i against allies of ( the cooperative federation (0,..., NPV, NPV,0,...,0), ove to Step (E). C. To select the next doinator k fro the cooperative federation to join into the negotiation gae, repeat Step (C) until Q = 0, ove to Step (E). jt D. To select the next doinator k fro the cooperative federation to join into the negotiation gae, repeat Step (C~D) until all the cooperative allies are selected as doinators, ove to Step (F). E. To select the next doinator according to the list of gae players to establish the negotiation gae, repeat Step (A~F), until all the gae list players are selected as doinators to coplete the construction of return atrix. (1) Look for gae equilibriu optial solution A. Fro the established copensation atr4ix, to select the gae cobination of axiu copensation as the starting point fro the doinator i (starting with 1). B. To proposed the corresponding cooperative strategy, and copare whether the copensation of the doinator k is the optial and acceptable cooperative strategy. If it is, ove to Step (D); otherwise, the doinator k proposes the suggested cooperation cobination and ove to Step (C). C. According to the cooperative strategy proposed by the doinator k as shown in Step (B), the doinator is to evaluate whether to accept or not, if so, ove to Step (D), otherwise, the doinator is to propose the next new cooperative strategy according to the suggestion of the doinator k before oving to Step (B). If the optial cooperative strategy cannot be found after continuous negotiations, this stage will be ended and ove to Step (E). 107

10 International Journal of Coputer Science & Inforation Technology (IJCSIT) Vol 6, No 1, February 2014 D. According to the obtained optial cooperative strategy, to select the next doinator k, repeat Steps (B~C). After all the cooperative allies are negotiated, the final optial cooperative strategy is the optial solution, and ove to Step (E). E. In order, to select the next doinator i, repeat Steps (A~D) until all the players on the list have copleted the negotiations to get the optial solution as the equilibriu solution of the negotiation gaes. F. To select fro all optial strategy cobinations the one of the highest total copensation as the optial solution.. According to the obtained optial cooperative strategy, to carry out order and calculate the order purchase. 4. ESULTS AND DISCUSSION According to the aforesaid 3C collaborative replenishent anageent odel and joint procureent decision odel negotiation algorith, we analyzed the sales data of the case chain restaurants in recent one year. The data of the first nine onths are the training data for product inventory odel and the sales data of the last three onths are the testing data for the product inventory odel. We applied the odel in the analysis of the ost representative, perishable packages of fresh eet category of food aterials of high unit price and large consuption. The three aterials are all independent aterials, the data and coputational results are as shown below: Table 1. Product TOP Data and BOM Table BOM Product TOPP Material A (g) Material B (piece) Material C (slice) A package 1, B package 1, C package 3, D package E package F package 6, C(NTD) Coefficient of losses (U) 10% 1% 2% Table 2. Material Deand ate Table Material Material A (g) Material B (piece) Material C (slice) Total MP 8,452,200 5,705 3,222 BOMM 5,871,600 4,125 1,429 Inv worst 2,113, ,276 96,657 2,551,983 Inv best 1,467, ,500 42,867 1,758,267 COMI Table 3. ABC Analysis and Order Model Selection Material Material A (g) Material B (piece) Material C (slice) ABC categorization A B C order policy regular quantitative order odel regular quantitative order odel S-s order control syste 108

11 International Journal of Coputer Science & Inforation Technology (IJCSIT) Vol 6, No 1, February 2014 Table 4. Order Policy Analysis Material Material A (g) Material B (piece) Material C (slice) Holding Cost (NTD) Purchase Cost (NTD) LT(year) TBP(year) EOQ OUT OP SS Note: SS=average daily sales volue (95% service level, estiated Shortage ate<2.5%). Table 5. Coparison of Ending Inventory Material A Material B Material C Ite BOM MP Holding Cost Filled ate Shortage ate 0.22% 2.85% Iproved ate 13.95% Holding Cost Filled ate Shortage ate 0.56% 0.88% Iproved ate 18.53% Holding Cost Filled ate Shortage ate 0.12% 0.48% Iproved ate 35.02% egarding A aterial, through coputation, we suggested that quantitative ordering inventory control syste should be applied. Copared to the traditional MP inventory control syste, it can be apparently found that the suggested quantitative ordering inventory control syste is better in ters of inventory replenishent and holding quantity control. Based on the statistical results as shown in Table 5, it can be apparently found that the suggested quantitative ordering inventory control syste has far lower holding costs as copared with the traditional MP inventory control syste. The perforance of inventory shortage control is far better than the traditional MP inventory control syste; the inventory replenishent rate can be ties. egarding B aterial, we suggested that quantitative ordering inventory control syste should be applied. Although it has poorer perforance in inventory shortage control as copared to the traditional MP inventory control syste, its overall perforance can be at 95% custoer service level. Based on Table 5, the suggested quantitative ordering inventory control syste is better than the traditional MP inventory control syste in inventory holding control, and the inventory replenishent satisfaction rate can be ties. egarding C aterial, we suggested that S-s order control syste should be applied. Based on Table 5, its perforance is better than the traditional MP in ters of inventory holding cost and shortage rate, and the inventory replenishent satisfaction rate can be ties. 109

12 International Journal of Coputer Science & Inforation Technology (IJCSIT) Vol 6, No 1, February Aount of Actual Consuption BOM Aount of Ending Inventory 0 MP Aount of Ending Inventory Date Quantity Figure 5. Coparison of Average Ending Inventories of Materials Through joint procureent and supply chain anageent coputation, iproved rate for the purchase of A aterial is 13.95%, 18.53% for B aterial and 35.02% for C aterial. It can be learnt fro Figure 5, the average end of ter inventories of the three food aterials have been significantly iproved. The syste after actual execution of the agent syste is as shown in Figure 6: Figure 6. Purchase Agent Syste The specific functions of purchase agent introduced as follows: 1. Interacting with user by interface agent. 110

13 International Journal of Coputer Science & Inforation Technology (IJCSIT) Vol 6, No 1, February By existing enterprise inforation systes interface counicate with original inforation syste. Databases and their associated functional odules described below: 1. Purchasing Knowledge Base: It contains the task decoposition ethods and the conflict resolution aong different Agents targets, intentions and Inference engine. To storage the purchase-related knowledge, through the interface and procureent controller interact with business knowledge base. 2. Agent Model Base: To store the use of agent odels and task identity base, include query agent, atching agent, negotiations agent, orders agent, and evaluation agent. (1) Query Agent: According to requireents by users inquiries, cobined with internal inforation base, knowledge base, as well as query the inforation needed by user. (2) Matching Agent: According to business purchase proposal and supplier proposal calculated atching level for autoatic atching, and subitted the higher degree of atching to negotiations agent. (3) Negotiations Agent: Based on the internal knowledge base, rule base interacting with supplies web service, coplete the negotiation. (4) Orders Agent: According to deand of users orders, aking orders to suppliers order processing web service online and iediate. (5) Evaluation Agent: According to the procureent process, cobined with corporate strategy related to the evaluation base policies (rules) for the evaluation of suppliers one by one, individually rated, and added to the enterprise knowledge base, ready for the later transaction reference. 3. Task Identity Base: To store the sets of task identity, cobined with agent odel base, decoposed procureent task into each sub-task, and assign to corresponding agent. 4. Inforation Delivery Mechanis Base: To store agents inforation delivery echanis; under the coordinated control of purchasing agent, often involved ultiple Agent to process of solving a proble, its requires coordination and dispatch of these Agent, and need to counicate between different agents, achieve interoperability inforation, inforation delivery agent is used to coplete this task. 5. Inference Engine: Based on decision odel, select the relative purchasing strategy (rules) fro the enterprise knowledge base and rule base, evaluation strategies (rule) to construct the procureent Knowledge base of e-procureent agent, then assigned the task of procureent under the procureent strategy (rules), according to the evaluation policy (rules) have been trading off copanies evaluate and rate each, adding to the knowledge base and update corporate procureent knowledge base and rule base through the interface or controller, provide a reference for future purchases. 5. CONCLUSION With the fresh food aterials of the chain restaurants as the exaple, we forecast product deand by using the coonality odel, production and delivery capacity odel and the odel of consuption with arket deand changes for classification and developent. Moreover, with the existence of dependencies between product deands as the deand forecast basis, we deterined the appropriate inventory odel accordingly. By integrating the dynaic strategy gae, we established the joint procureent decision odel negotiation algorith, and used the siulation case for analysis and verification to achieve the design of the dynaic negotiation optiization echanis for the joint procureent of food aterials. By using the proposed ethod for inventory anageent, we can indeed effectively reduce inventory quantity, inventory cost, and thereby enhancing the custoer service satisfaction level. 111

14 International Journal of Coputer Science & Inforation Technology (IJCSIT) Vol 6, No 1, February 2014 EFEENCES [1] eid Sith, (1980) The contract net protocol: High-level counication and control in a distributed proble solver, Coputers, IEEE Transactions on, Vol. 1000, No. 12, pp.1104~1113. [2] Marco Cantaessa, (1997) Hierarchical and heterarchical behaviour in agent-based anufacturing systes, Coputers in Industry, Vol. 33, No. 2, pp.305~316. [3] Sion Case, Nader Azari, Marcus Thint, & Takeshi Ohtani, (2001) Enhancing e-counities with agent-based systes, Coputer, Vol. 34, No. 2, pp. 64~ 69. [4] David J Teece, (2010) Business odels, business strategy and innovation, Long range planning, Vol. 43, No. 2, pp. 172 ~ 194. [5] Yu-Teng Chang, Chih-Yao Lo, Pin-Chang Chen & uei-chi Tang, (2011) "esearch on the Developent of Web-based Joint Procureent Inforation Systes", International Journal of Digital Content Technology and its Applications, Vol. 5, No. 9, pp. 134 ~ 144. [6] Moshe Kaspi & Meir J osenblatt, (1985) The effectiveness of heuristic algoriths for ulti-ite inventory systes with joint replenishent costs, International Journal of Production research, Vol. 23, No. 1, pp. 109 ~ 116. [7] Moshe Kaspi & Meir J osenblatt, (1991) On the econoic ordering quantity for jointly replenished ites International Journal of Production esearch, Vol. 29, No. 1, pp. 107~ 114. [8] SK oyal, (1974) Optiu ordering policy for a ulti-ite single supplier syste, Operational esearch Quarterly, Vol. 25, pp. 293~ 298. [9] Brown.. (1967) Decision rules for inventory anageent, Holt, einhart and Winston, New York, pp [10] SK oyal & AS Belton (1979) A siple ethod of deterining order quantities in joint replenishents for deterinistic deand, Manageent Science, Vol. 25, No. 6, pp. 604 ~ 604. [11] Edward A Silver, (1976) A siple ethod of deterining order quantities in joint replenishents under deterinistic deand, Manageent Science, vol. 22, No. 12, pp. 1351~ [12] Xiaofang Zhang & Wei Zhu, (2012) "The Study on Food Quality Supervision using Collusion ae Model", International Journal of Advanceents in Coputing Technology, Vol. 4, No. 4, pp. 283 ~ 293. [13] Yingxue Zhao, Shouyang Wang, TC Edwin Cheng, Xiaoqi Yang, & Zhiin Huang, (2010) Coordination of supply chains by option contracts: A cooperative gae theory approach, European Journal of Operational esearch, Vol. 207, No. 2, pp. 668 ~ 675. [14] Jinshi Wei & Hongjie Lan, (2011) "Establishing Food Traceability Syste based on ae Theory fro the Perspective of etailers", Advances Inforation Sciences and Service Sciences, Vol. 3, No. 6, pp. 107 ~ 114. [15] Ai-Ping Li, Yan Jia, & Quan-Yuan WU, (2007) "A Study on Organizational Knowledge in Multi-Agent Syste", Journal of Convergence Inforation Technology, Vol. 2, No. 1, pp. 61 ~ pp.65. [16] Yanzhong Li, Zhixia Jiang, Pinchao Meng, Weishi Yin, & Jun Li, (2013) "A Collaborative Optiization Method for Solving Multi-objective Prograing Proble", International Journal of Advanceents in Coputing Technology, Vol. 5, No. 1, pp. 809 ~ 817. [17] Scott A DeLoach, Mark F Wood, & Clint H Sparkan, (2001) Multiagent Syste Engineering, International Journal on Software Engineering and Knowledge Engineering, Vol. 11, No. 3, pp. 231 ~ 258. [18] Michael Wooldridge, Nicholas Jennings, & David Kinny, (2000) The aia Methodology for Agent-Oriented Analysis and Design, Journal of Autonoous Agents and Multi-Agent Systes, Vol. 3, No. 3, pp. 285 ~ 312. [19] Chun-Ta Lin, Chih-Yao Lo, & Chia-Hsing Lin, (2008) Apply 3C Theory in Spare parts Manageent in Mobile Phone Industry, Journal of Applied Science, Vol. 8, pp ~ [20] Ayako Kawai, (2012) "Effect of Structural Coordination on Supply Chains Controlled by Manufacturing Planning and Control Systes", International Journal of Engineering and Industries, Vol. 3, No. 3, pp. 84 ~ 93. [21] Jiawang Xu, (2013) "Multi-objective Operation Model for Supply Chain with Uncertain Prices Based on Fuzzy Sets and obust Optiization", Advances Inforation Sciences and Service Sciences, Vol. 5, No. 2, pp. 320 ~