WEB-BASED MULTI-ATTRIBUTES ANALYSIS MODEL FOR MAKE-OR-BUY DECISION 8th ISAHP 2005, Hawaii, University of Hawaii Honolulu, July 8-10, 2005 Prof. Heung Suk Hwang, Department of Business Management , Kainan University, Taiwan Tel : +886-3-341-2500 ext. 6088 e-mail : hshwang@mail.knu.edu.tw Kainan University 1 Contents 1. Introduction 2. Properties of Make-or-Buy Decision Problem 3. Three-step Approach of Decision Alternative Analysis, Project Risk Analysis Models 4. Model Application to Cellular Manufacturing System 5. Summary and Conclusions Kainan University 2 1. Introduction ☞ Developed an internet/intranet-based solution builder (Solution Builder2004) for a three-step multi-attribute decision support system for School Food Service System ☞ Web-based Three-step Decision Support System Using Multiattribute Analysis Method 1) brainstorming for the idea generation, 2) fuzzy-AHP( fuzzy analytic hierarchy process) as a multi-attribute structured an analysis method , 3) aggregation logic model to integrate the results of individual analysis ☞ We applied this decision support system to the make-or-buy decision problem for school foodservice system considering the multi-attributes in the decision making. ☞ Web-based GUI-Type Program is developed and demonstrated in internet/intranet-based decision problem. Kainan University 3 Web-based Decision Support System Group-Joint Work Web-based Integrated Decision Support System Information System Internet/Intranet Web-based Integrated Decision Support System Kainan University 4 Step 1 : Individual Evaluation of Alternatives 1) Brainstorming to Generate Alternatives and to Define the Performance Factors 2) Evaluation of Alternatives Using AHP and Fuzzy AHP methodologies Step 2 : Integrate the Individual Analysis - Heuristic Model 1, 2 - Fuzzy Set Priority Method Step 3 : Application, Resource Allocation Model - LP formulation using AHP weighted value - Developed Computer Program - Web-based Internet/Intranet Solution Builder - GUI-type Program - Integrated decision support system Figure 2 . Three-step Approach of the Evaluation Model Kainan University 5 2. Properties of Make-or-Buy Decision Problem ☞ Properties and issues responsible for differentiating one type of make-or-buy decision problem . - What backgrounds trigger a make-or-buy decision problem? - What factors could be considered in make-or-buy decision problem ? - Along which dimensions should make-or-buy decision problem be categorized ? Kainan University 6 Table 1. Major factors Influencing make-or-buy decision problem (by literatures) Performance Measure Criteria • Cost • Quality • Delivery speed • Delivery reliability • Volume flexibility • Product flexibility Kainan University Examples of measurement Parameters - Total unit cost - Internal failure cost-scrap, rework, rejected - Delivery lead time - Percentage of on-time delivery - Average volume fluctuation - Number of component substitutions made over a given time period. 7 Make Low High Flexibility Control High Low Full Ownership Partial Ownership Retainer Short Term Contract Spot Market Buy Figure 3. Range of Source Structure of make-or-buy decision problem Kainan University 8 Table 2. Example of performance Evaluation of make-or-buy decision problem Item Manufacturing Technology Out Source Risk Managerial Issues Financial Issues Operational Issues Major Factors - Importance of technology for competitive advantage - Maturity of technology - Technology uncertainty - Probability of future improvements - Appropriation risk - Technology diffusion - End-product degradation - Benchmarking - Workforce stability - Complexity level in planning, control, or supervision - Assurance and reliability of supply - Benchmarking - Cost - Investment - Return on investment - Manufacturing capability - Quality - lead time - Volume uncertainty 9 3. Three-step Approach of Decision Alternative Analysis Model Step 11 :: Brainstorming Stochastic Set 단계 - Generate Alternatives and ? 적정보급센터의 소요Factors - Define the Performance 보급센터의 위치결정 - Relationship Between Factors 적정 보급지원수준결정 Hierarchy Step Secter-Clustering Model Process 단계22 :: AHP,Analytic - Construction Evaluation Structure - Evaluation of Alternatives Using AHP 영역활당 and보급지원 Fuzzy AHP methodologies Zone-Based -- Visual 시각화Program • GYI-Type GUI-Type ?•프로그램개발 SW Developed 프로그램개발 ?•사용자 사용자위주의 위주의 Customer Responsive 프로그램개발 프로그램개발 Web -- Web-based 기반의 통합화 Network Netork • System System System System • 확장성 Flexibility , 확장성 • Usability 활용성 활용성 Model Step 33 :: Aggregating GA-VRP Model -Evaluation - 개별 of Alternatives Using AHP and Fuzzy set ranking methodologies - 종합우선순위산출 - Prioritize the Prioritized Sets 운송 Mode 의 선정 Fig 3. Web-based Integrated Decision Model 10 3.1 Brainstorming ☞ Construct decision structure and Derive out the evaluation alternatives - the group decision ideas, the creative ideas ☞ we used a brainstorming method and developed a GUI-type program ☞ To create the ideas of project evaluation alternatives and methods for decision support system analysis, ☞ we construct decision structure using the brainstorming file in the internet/intranet–based environment 11 3.2 Fuzzy -AHP Method ☞ The concepts and rules of fuzzy decision making provide us with the necessary tools for structuring a decision from a kind of information. ☞ From the Shannon's summed frequency matrix for complementary cells, - ☞ an additional fuzzy set matrix was made by considering Aij = 1 – A ji for all cells. The fuzzy matrix complement cell values sum to 1 and fuzzy set difference matrix is defined as follows : R - R T = U(A, B)-U(B, A), if U(A, B) > U(B, A), = 0 otherwise where, for U(A, B) quantifies, A is preferable to B. Kainan University 12 Five Steps Fuzzy AHP : To obtain fuzzy preferences, the following five steps were considered: Step 1 : Find the summed frequency matrix ( using Shannon method ) Step 2 : Find the fuzzy set matrix R which is the summed frequency matrix divided by the total number of evaluators Step 3 : Find the difference matrix R - R T = U(A, B)-U(B, A), if U(A, B) > U(B, A), X = 0 otherwise where, for U(A, B) quantifies, A is preferable to B. 1.ColA Step 4 : Determine the portion of each project that is not dominated as follows : ND = 1 - max( ,… ) X 1.ColA X, 2.ColA X n,.ColA AColA Step 5 : The priority of the fuzzy set is then the rank order of XND values with a decreasing order. Kainan University 13 An example is shown as follows : 0 .0 0 .2 R= 0 .4 0.4 0 .0 R T = 0.8 0 .6 0.6 0 .0 0 .0 T RR 0 .0 0.0 Kainan University 0 .6 0 .0 0 .0 0 .4 0 . 1 0 .0 0 .6 0 .6 0 .4 0.0 0 .2 0 .0 0 .4 0.1 0 .4 0 .6 0 .0 0 .4 0 .0 0 .4 0 .6 0 .0 0.8 0 .0 0 .6 0 . 2 0 . 2 0 .0 0 . 0 0 . 0 0 .1 0 . 0 0 . 0 0.2 0.2 0.0 14 ND XB ND XA ND XC ND XD = 1 - Max(0.0) = 1 - 0.0 = 1.0 = 1 - Max(1.0) = 1 - 1.0 = 0.0 = 1 - Max(0.2) = 1 - 0.2 = 0.8 = 1 - Max(0.2) = 1 - 0.2 = 0.8 The fuzzy set priority score : 1.0 > 0.0 > 0.8 > 0.8 and the alternative priority : Kainan University A > C > D > B. 15 3.3 Integration of Individual Evaluation ☞ For the integration of the results of individual evaluations, prioritized sets, we used two Heuristic models 1, Model 2 and Fuzzy set priority method 1) Heuristic Model 1 : - For example of the Heuristic Method 1, a sample result with - N = 5 evaluators and M = 3 alternatives is given as : Evaluator 1 : B > A > C, Evaluator 2 : B > C > A, Evaluator 3 : C > A > B, Evaluator 4 : C > B > A, Evaluator 5 : C > B > A Kainan University 16 ☞ Heuristic Method 1 rank order is given by C(0.467) > B(0.400) > A(0.133). Kainan University 17 2) Heuristic Model 2 : - The evaluator frequency matrices were added to form a summed frequency matrix - Then, the preference matrix was developed by a comparison of the scores in the component cells(A, B versus B, A). - If the A, B value equals B, A, then each component cell in the matrix is given by 1/2. On the other hand if the A, B value is greater than the B, A , then A, B is given by one and B, A cell of the preference matrix is given by 0. ☞ By applying the Heuristic Model 2 to the same example of Heuristic Method 1, the result is given by C(0.450) > A(0.392) > B(0.158) . Kainan University 18 3) Fuzzy Set Priority Method . The fuzzy matrix complement cell values sum to 1 and fuzzy set difference matrix is defined as follows : R-RT = U(A, B) - (B, A), if U(A, B) > U(B, A), = 0, otherwise To obtain fuzzy preferences, following five steps are considered : Step 1 : Find the summed frequency matrix (using heuristic method 2) Step 2 : Find the fuzzy set matrix R which is the summed frequency matrix divided by the total number of evaluators Step 3 : Find the difference matrix R - RT = U(A, B) - U(B, A), if U(A, B) > U(B, A), = 0, otherwise where, for U(A, B) quantifies, A is preferable to B. Step 4 : Determine the portion of each part Step 5 : The priority of the fuzzy set is then the rank order of values in decreasing. The sample problem result by fuzzy set priority method is given by C(0.492) > B(0.387) > A(0.121). Kainan University 19 3.4 In ternet /intranet Based Solution Builder for Decision Support System ☞ Developed a solution builder using GUI-type Simulation Software. ☞ Three steps of this solution builder. 3-step Algorithm for Optimal Solution Brainstorming AHP, Fuzzy--AHP Aggregate Priorities Figure 2. 3-step approach of Decision Support System Kainan University 6 20 Server Protocol Encoding Network Internet/Intranet Protocol Decoding Client Figure 4. Client and Server in Decision Support System Kainan University 21 Start Start Input Data (Sub list) Consistency Test Data Transform Output Frequency Matrix Generate Output Output Output Heuristic Borda Rank 1 Rank order Order Fuzzy Priority Consistency Test Summation Frequency Matrix Fuzzy Matrix Frequency Matrix Shannon Heuristic 2 Copeland Rank Order Rank Order Output STOP STOP Fig 6. Schematic Flow Diagram of the Proposed Model Kainan University 22 The GUI-type program of Solution Builder-2001 Edit File New File Show Edit Initial Size Add Enlargement Reduce Delete Open Tool Bar Move Close Copy Save Standard Search Node Insert Item Status Bar Save in Other Name Select All Project Information Line Up Character Line up Left - line Encoding Sort Right - line Printer Setup AHP Upper - line Integrating Ranking Change Bottom - line Pre-Show Edit Levels Centering Print Add Level Centering Vertical E-Mail Sending Delete Level Exit Brainstorming AHP Integrate Convert Brainstorming File into AHP File AHP Convert AHP File into Integrating Ranking File Aggregating Ranking Method Kainan University 23 Aggregating Priority AHP Compute Basic Information Network Server Setup Tool Copy in Clipboard Partial Mode Heuristic 1 Ideal Mode Heuristic 2 Save in Image File Perf. Sensitivity Fuzzy Set Priority Option Dynamic Sensitivity Trend Sensitivity Node Information Client Setup Save in HTML Help Index Homepage Solution Builder? Figure 5. Main-program of Solution Builder 2001 Kainan University 24 ☞ We used a brainstorming method and developed a GUI-type program Edit File New File Open Close Save Save in Other Name Project Information Encoding Printer Setup Pre-Show Print E-Mail Sending Exit Show Edit Initial Size Add Enlargement Delete Move Copy Insert Reduce Tool Bar Standard Search Node Item Status Bar Character Select All Line up Line Up Sort Left - line Change Upper - line Right - line Bottom - line Centering Centering Vertical Kainan University 25 4. Make-or-Buy Decision Analysis in Cellular Manufacturing Syatem ☞ Applied to Special decision problems; multi-objective, multi-criterion, and multi-attributes structures for Cellular manufacturing system 1) Make-or-buy decision making, 2) Determine the weighted value of each decision factors, 3) Resource allocation in manufacturing process, Kainan University 26 4.1 Cellular Manufacturing System ☞ Generally, the cellular manufacturing system uses many kinds of machines and tools ☞ manufacturing process is a little bit complicated than conventional production system ☞ In this study we used an oil pan manufacturing cell ☞ produces oil pan by 120 lot size, two workers, and CNC machine: - milling machine, - boring machine, - multi-spindle and drilling, ☞ CNC cell produces oil pan by 120 lot sizes. 27 ☞ Sample Example : Oil pan manufacturing cell layout CNC Drilling, Multi-spindle, And tapping Machine Governer Assembly Area Finished Oil pans Oven Casting (Raw Material) Dong-eui University, Korea CNC Boring Machine CNC Milling, Machine 28 Figure 5. Sample output of AHP Structure (Oil Pan Manufacturing Cell) Kainan University 29 AHP Structure Figure 6. Sample Output of AHP structure of Cellular manufacturing System 30 Level 1 A1 Final Object (Final Object) 0.74 Make Level 2 in house B1 (Acq. Method) 0.38 Level 3 (Alternative) P1 Proj. 1 B1 B2 B2 0.19 0.39 0.58 B2 0.26 P2 Proj. 2 0.29 0.32 0.16 0.20 0.06 Partial Make Buy tech import 0.19 P3 Proj. 3 0.21 0.21 0.15 B3 outsourcing 0.12 P4 Proj. 4 0.05 P5 Proj. 5 0.29 0.06 0.07 0.02 0.04 0.04 31 Integration of Individual Evaluations : Using the Heuristic 1, Heuristic 2, AHP, and Fuzzy Set Ranking methods, we integrated the results of the individual reviewers as following, where, B1: make in house not outsourcing, B2: partial make in house and partial out sourcing for technology, B3: all outsourcing, P1, ···, P5 : cellular manufacturing alternatives Table 4. Results of Integrated Priority Majority Rule used Priority by Alternative Methods Priority by Alternatives 1. Heuristic Model 1 B1 (0.70), B2 (0.18), B3 (0.12) P1 (0.29), P2 (0.30), P3 (0.18), P4 (0.15), P5 (0.08) 2. Heuristic model 2 B1 (0.73), B2 (0.23), B3 (0.05) P1 (0.36), P2 (0.27), P3 (0.13), P4 (0.15), P5 (0.09) 3. Fuzzy Set Ranking Method B1 (0.74), B2 (0.20), B3 (0.06) P1 (0.38), P2 (0.26), P3 (0.19), P4 (0.12), P5 (0.05) 32 5. Resource Allocation in Cellular Manufacturing System ☞ Using the AHP weighted value in resource allocation of manufacturing works for Cellular manufacturing system ☞ For the budget allocation problem for this cellular manufacturing works (alternatives) using the weighted values of level 2, we formulated as following optimization problem. Kainan University 33 n m i 1 j 1 Max Wij X ij Kainan University 34 35 Formulation: Max Z = 0.19X11 + 0.29X12 + 0.21X13 + 0.29X14 + 0.02X15 + 0.39X21 + 0.32X22 + 0.21X23 + 0.06X24 + 0.04X25 + 0.58X31 + 0.16X32 + 0.15X33 + 0.07X34 + 0.04X35 s.t. 11000X11 + 9000X12 + 12000X13 + 8000X14 + 7000X15 ≤ 25000 4000X21 + 5000X22 + 6000X23 + 5000X24 + 3000X25 ≤ 18000 4000X31 + 3000X32 + 5000X33 + 2000X34 + 1000X35 ≤ 11000 where, Xij = 0, 1 ∀i, j 36 37 6. Summary and Conclusion ☞ a three-step approach of web-based make-or-buy decision model for multi-structured decision support system: 1) Brainstorming to define the alternatives and performance evaluation factors, 2) Individual evaluation of the alternatives using fuzzy-AHP, heuristic and fuzzy set reasoning methods, and 3) Integration of the individual evaluations using majority rule method. ☞ Developed a systematic and practical program ☞ The model was applied to a cellular manufacturing system problem for the purpose of comparative validation. ☞ The results of various multi-structured decision support examples for make-or-buy decision analysis and also resource allocation problems are shown By the sample results, the proposed model is a good method for the performance evaluation of multi-attribute and multiple goals for make-or-buy decision problems. ☞ 38 Thank You Kainan University, Taiwan Prof. Heung-Suk Hwnag Kainan University 39