Chapter 4 Methodology
4.3 The Research of Fuzzy Cognitive Map (FCM)
4.3.5 The Processing of FCM
All the experts are consulted with their experience and knowledge, and ultimately evaluated on an appropriate numerical scale. At each simulation step of the FCM, the value of concepts is calculated according to equation (1). Where Ci(tn+1) is the value of concept Ci at step tn+1, Ck(tn) is the value of concept Ck at step tn, and wki(tn) is the weight of the interconnection from concept Ck to concept Ci.
S(x) is a threshold function that squashes the result of the multiplication in the interval [0, 1]. The logistic signal function has been used to transform to an S-shaped curve as Figure 4.4.
x
11ex
S
(7)
Figure 4.4 The logistic signal function
In general, as observed in Figure 4.4, the input is the setting in the interval [-10, 10] and also sets the constant αfor 4. Then, after calculating the logistic signal function, we can get
41
the value in the interval [0, 1].
From Equation 1, k doesn‘t equal to i, it means that concept evaluates from the standpoint of self-interest. In order to clarify, whether the self-evaluation for the concept is good or not, we conducted tests basing on equation (8). Where
j represents the weight of other concepts, and (1-
j) represents how they evaluate themselves. The result indicated that the more you evaluate on just the concept, the pace in system improvement will be more slowly.
n
j indicates that the system was not stability at the beginning; such a partial explanation for this may lie in the fact that the product category (D) is not stable. To increase the order of 24-hour delivery (A) and improve the ability to achieve 24-hour delivery (E), we need about 10~25 workdays to adjust. After 30 workdays, we find that the Ability to achieve 24-hour delivery is steady at 0.88 in the long run.Figure 4.5 The output of FCM (β=0.5)
Figure 4.6 illustrates the concept consider about themselves for (1ｊ) equal to 0.95.
Compare to Figure 4.5, we find out the system needs 36 runs to achieve the goal. The result indicated that the more you consider on each factor independently, the system will be slower
42
to improve. Considering the overall promotion significantly accelerates the improve process.
Figure 4.6 The output of FCM (β=0.05) 4.3.6 The Output of FCM (β=1)
After the data collection and processing through equation (1), the original output from the proposed FCM is as follow and depicted in Table 4.8.
Table 4.8 The runs of calculation and their respective value
Run A B C D E F G H I J
0 0.89 0.89 0.67 0.71 0.83 0.60 0.77 0.51 0.46 0.66 1 0.91 0.72 0.96 0.78 0.53 0.27 0.99 0.49 0.86 1.00 2 0.84 0.83 0.90 0.65 0.63 0.18 0.98 0.45 0.93 1.00 3 0.77 0.83 0.90 0.70 0.69 0.17 0.97 0.39 0.92 1.00 4 0.82 0.81 0.89 0.70 0.90 0.20 0.96 0.37 0.93 1.00 5 0.84 0.82 0.93 0.69 0.90 0.33 0.97 0.32 0.93 1.00 6 0.86 0.79 0.94 0.70 0.87 0.32 0.97 0.37 0.93 1.00 7 0.85 0.80 0.94 0.68 0.88 0.33 0.97 0.37 0.93 1.00 8 0.85 0.80 0.94 0.69 0.88 0.33 0.97 0.37 0.93 1.00 9 0.85 0.80 0.94 0.69 0.88 0.33 0.97 0.37 0.93 1.00 10 0.85 0.80 0.94 0.69 0.88 0.33 0.97 0.37 0.93 1.00 11 0.85 0.79 0.94 0.69 0.88 0.33 0.97 0.37 0.93 1.00 12 0.85 0.79 0.94 0.69 0.88 0.33 0.97 0.37 0.93 1.00 13 0.85 0.79 0.94 0.69 0.88 0.33 0.97 0.37 0.93 1.00 14 0.85 0.79 0.94 0.69 0.88 0.33 0.97 0.37 0.93 1.00 15 0.85 0.79 0.94 0.68 0.88 0.33 0.97 0.37 0.93 1.00
43
It has to be stressed that the study observes the rule that, if there is no sudden accident to the case company, it will survive as time goes on. The concepts C, G, I, and J run after 3 runs and their results observed are close to the range [0.93, 1]. It shows that relationship with suppliers (C), resilience of safety stock (I), and stable operation of warehouse (J) experience almost no changes in the long run. To summarize the salient features of the analysis, several findings should be of interest. Figure 4.7 illustrates what has been transpired in that model.
We could observe that the value of each concept will vibrate when the vibration is stable3. In addition, it is important to emphasize that the incorporated concepts need to take time to adapt to real world in the scheme of research design. Each run is assumed to be 5 workdays;
means that PChome needs about 5 workdays to respond and address all the necessary condition changes.
Figure 4.7 The output of proposed FCM
The first step is test, to explain that system had no stability at the first 10 workdays. Such a partial explanation for this may point to the fact that the product category is not stable. The
3 The results of the t1 t 0.001means the value is stable.
44
second step is adjustment, to increase the order of 24-hour delivery and improve the ability to achieve 24-hour delivery, we need about 10~25 workdays to adjust and adapt. Finally step is when 30 days after the system stabilizes; we should be able to witness that the ability to achieve 24-hour delivery is steady at 0.88 in the long run.
The results reflected in Figure 4.7 indicate that:
(1) Logistics performance (concept B) was decreased at the first run, and after 5 workdays it gradually increases, then drop back down a bit about 25 workdays after. This would more likely explain that logistics performance was not stabilized at the beginning.
After 30 workdays the logistics performance was closes to 0.79 representing that, logistics performance will have almost no changes in the long run.
(2) Product category (concept D) increased at the first and decreased at the second runs.
A partial explanation for this may lie in the fact that the product category is not stable. Then it gradually increases, and after 15 workdays the product category is at a steady within the range of [0.68, 0.69].
(3)Ability to achieve 24-hour delivery (concept E) at the first run decreased, followed by an immediate sharp increase between 5 to 20 workdays. Then the ability of achieving 24-hour delivery gradually declines for a duration of 25 to 30 workdays. We can conclude that PChome Company‘s major goal is to increase the order of 24-hour delivery and improve the ability to achieve 24-hour delivery. It is glad to see that the ability to achieve 24-hour delivery is steady at 0.88 in a longer term.
(4)Lack of ability to develop information system (concept H) decreased until 6 runs, and dropped to a range [0.32, 0.37] after 25 workdays. It may be due to PChome company is not a computer design company, so they may take a long time to resolve this deficiency and improve it. One possible solution is considering outsourcing to other information industry, but
45
PChome Company insists of having faith in their information systems.
(5)Time window problem (concept F) closes to 0.20 until 5 runs, and also represents time window problem experiencing slight increases from 25 to 30 workdays. The time window problem is stabilized to relatively low about 0.33 after 30 workdays.
If the orders for 24-hour delivery service remain high, relationship with suppliers will be good, and the lack of ability to develop information system will decrease. Thanks to the decreased time window problem, the ability to achieve 24-hour delivery will stay at the desired high level.
46
Chapter 5 Scenario Analysis
5.1 Scenario Introduction
In Taiwan, portal sites provide consumers with shopping on-line service in an electronic store and picking up goods within 24 hours. Based on the research background and motivations, the results revealed that order of 24-hour delivery service (A), logistics performance (B), relationship with suppliers (C), product category (D), ability to achieve 24-hour delivery (E), time window problem (F), information system (G), lack of ability to develop information system (H), the resilience of safety stock (I), and stable operation of warehouse (J) can be used to construct the system relationship model.
5.1.1 Scenario 1 - Order of 24-hour delivery increase
PChome Company‘s major goal is to increase the order of 24-hour delivery and improve the ability to achieve 24-hour delivery. As mentioned above, we get the rating of concept state we put it at the bottom of the Table 4.6, it represents the current state and it means the order of 24-hour delivery is good (A=0.89) now. Although we all think of the goal is to increase order of 24-hour delivery and improve the ability to achieve 24-hour delivery, a number of intervention conditions have also been considered. Such us holiday order, typhoon order and so on. Hence, it is assumed that order of 24-hour delivery service increase suddenly.
Therefore, the research set the initial state value of order of 24-hour delivery service (concept A) at 1.0, to observe the differences between the original output and the adjusted ones.
After setting the initial state value of order of 24-hour delivery service (concept A) to 1.0, the original input of proposed FCM is as follows Table 5.14.
4 The table filled in ―( )‖means negative number.
47
Table 5.1 The input data
A B C D E F G H I J
A 0.00 0.58 0.65 0.00 0.00 0.00 0.73 0.00 (0.48) 0.59 B 0.00 0.00 0.00 0.50 0.00 (0.83) 0.00 0.00 0.00 0.00 C 0.58 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.51 0.00 D 0.69 0.00 0.00 0.00 (0.33) 0.00 0.15 0.00 0.60 0.58 E 0.27 0.00 0.48 0.00 0.00 0.73 0.00 (0.30) 0.00 0.00 F 0.00 (0.46) 0.00 0.00 (0.40) 0.00 0.25 0.39 0.00 0.35 G 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.20 0.05 H (0.20) 0.00 0.00 0.00 0.00 0.00 0.28 0.00 (0.10) 0.00 I (0.48) 0.00 0.00 0.00 0.38 0.10 0.00 0.00 0.00 0.53 J (0.30) 0.00 (0.30) (0.20) 0.49 (0.25) 0.00 0.00 0.00 0.00 Current state 1.00 0.89 0.67 0.71 0.83 0.60 0.77 0.51 0.46 0.66
Figure 5.1 The output of scenario 1 The results reflected in Figure 5.1 indicate that:
(1) Findings of this study points to the fact that, if there are no sudden and drastic changes within the parameters and variables from the case company, all the findings so far would hold true. The concepts C, G, I, and J run after 3 runs close to the range [0.93, 1].
(2) Logistics performance (concept B), ability to achieve 24-hour delivery (concept E) and time window problem (concept F) rapidly decreased at the first run.
48
(3) Lack of ability to develop information system (concept H) did not decrease until the 6th run, and dropped to a range [0.37, 0.38] after 35 workdays.
Its orders from 24-hour delivery service will remain high, relationship with suppliers will still be good, and the lack of ability to develop information system will increase. Thanks to the decreased incidents of time window problem, the ability to achieve 24-hour delivery will stay at high level.
49
5.1.2 Scenario 2 - Sudden Dip of Time Window Problem
Time window is a common form of time constraint extensively evaluated in the literature.
The results reflected in Figure 4.4 indicate that time window problem gradually declines. The balance value is closes to 0.33, and represents time window problem will drop a little bit about 25 workdays later. Although suppliers (or customers) always request us send the good in some time interval what they wanted, a number of conditions have also been considered.
The dipping of time window problem is possibly due to the weakness of the request of suppliers (or customers). Hence, we assumed that PChome Company‘s time window problem precipitates to the lowest point suddenly, and we respond by changing the output data for 0 on the runs of 7 to see the differences between the original output and the adjusted ones. The input of proposed FCM is as follows Table 5.2.
Table 5.2 The output data of scenario 2
Run A B C D E F G H I J
50
Figure 5.2 The output of scenario 2 The results reflected in Figure 5.2 indicate that:
(1) It has to be stressed that the study observes if there is only one sudden and unpredictable incident to the case company, it is assumed that PChome Company‘s time window problem falls straight to the lowest point on the run 7. The concepts C, G, I, and J run after 3 runs close to the range [0.93, 1].
(2) Logistics performance (concept B), ability to achieve 24-hour delivery (concept E) and time window problem (concept F) decreased at the first run. The sudden incident has led to the highest point of logistics performance and ability to achieve 24-hour delivery.
(3) Lack of ability to develop information system (concept H) increased until 6 runs, and the sudden incident has led to an abrupt dip on run 8.
Thanks to the decreasing in time window problem, the ability to achieve 24-hour delivery will remain at high level.
51
5.2 Scenario analysis 5.2.1 Scenario 1 analysis
Figure 5.3 illustrates the value difference between the adjusted scenario and the original scenario according to the scenario runs. Each run is assumed to last 5 workdays after the consultation with experts.
Figure 5.3 The output of the original FCM and adjusted scenario 1
(1) Order of 24-hour delivery service (concept A) decreased sharply about 20 workdays.
(2) Logistics performance (concept B) decreased at the first run about 5 workdays.
Compared with the original scenario, when the order of 24-hour delivery became steady, the logistics performance became steady, too.
(3) Product category (concept D) increased at the first run, and the product category declined to 0.65 between 5 to 15 workdays. Compared with the original scenario that represents the order of 24-hour delivery increase will reduce product category. After 20 workdays the product category is steady at 0.68.
52
(4) Ability to achieve 24-hour delivery (concept E), the output of adjusted scenario can be equated with the original scenario, that is to say the system can help PChome to achieve 24-hour delivery.
(5) Time window problem (concept F) closes to 0.20 until 4 runs, and also represents time window problem slight increases from 25 to 30 workdays. Compared with the original scenario, it is something different from 5 to 10 workdays that represents the order of 24-hour delivery increase will trigger more time window problem. The time window problem is also steady at low status about 0.33 after 30 workdays.
Figure 5.4 The difference between the scenario 1 changes
The following changes should be noted: The logistics performance (concept B) and product category (concept D) decrease for about 10 and 15 workdays each. The increase of relationship with suppliers (concept C) starts after the 10th workday and ends after 15th workday. The statuses of the resilience of safety stock, time window problem, and lack of ability to develop information systems are comparatively diminishing at initial stage and stay the same with the original output in a longer term.
53
5.2.2 Scenario 2 Analysis
Figure 5.5 illustrates the value difference between the adjusted scenario 2 and the original FCM according to the runs from scenarios. Each run is assumed to be 5 workdays after the consultation with experts.
Figure 5.5 The output of the original FCM and adjusted scenario 2
(1) Order of 24-hour delivery service (concept A): The curve adjusted scenario 2 is similar to original curve.
(2) Logistics performance (concept B): Compared with the original scenario, the sudden incident has led to the highest point of logistics performance after 5 workdays.
(3) Product category (concept D): Compared with the original scenario, the sudden incident has led to the highest point of product category after 10 workdays, and after 15 workdays the product category is steady at 0.69.
(4) Ability to achieve 24-hour delivery (concept E): The sudden incident has led to the highest point of ability to achieve 24-hour delivery after 5 workdays. Without this, the output
54
of adjusted scenario can be equated with the original scenario, that is to say the system can help PChome to achieve 24-hour delivery.
Figure 5.6 The difference between the scenario 2 changes
The following changes would be noticed: Order of 24-hour delivery service (concept A) is similar to that of the original curve. An interesting finding is that has led to the highest point of order of 24-hour delivery service (concept A) after 5 workdays. The logistics performance (concept B), and ability to achieve 24-hour delivery (concept E) also increase for about 5 workdays each. This sudden incident has led to the highest point of product category (concept D) after 10 workdays. The statuses of the resilience of safety stock (concept I), the relationship with suppliers (concept C), information system (concept G) and lack of ability to develop information systems (concept H) are comparatively diminishing at the initial and be the same with the original output in the long run.
55
Chapter 6 Conclusions and Suggestions
6.1. Conclusions
In Taiwan, the service of shopping on-line in an electronic store and picking up goods at home could be quite common and popular these days. Based on the research background and motivations, the research issues tend to point out the relationships to the 24-hour delivery, and they are as follow:
Issue 1:
Through the literature reviews and in-depth interviews as well as discussions with experts, this research constructs 10 important concepts and 13 criteria for the achievement of 24-hour delivery service. Then, proceed to construct the system relationship model. The reported in this paper have demonstrated that these two systems can be practically implemented, and hope to applicative to the other company. In conclusion, SM and FCM could be used for systematic studies both as an instruction tool and research tool.
Issue 2:
From SM, we find out that order of 24-hour delivery service (A) and time window problem (F) are major components of 24-hour delivery system; that material entities are order of 24-hour delivery service (A), logistics performance (B) and resilience of safety stock (I).
The high scores within the last three criteria of the category of information remind us that information flow exchange is also taking into the account of the external factors during the process of policy formulation. Lack of ability to develop information system (H) for PChome company, the more likely explanation may be due to PChome company is not a computer design company; the possible solution is considering outsourcing to other information industry.
Issue 3:
The findings of FCM lead to a number of implications. The order of 24-hour delivery
56
service is strongly affected by the relationship with suppliers. Relationship with suppliers is a critical factor that impacts whether how many the suppliers would transfer the orders to 24-hour delivery of PChome. Keeping good relationship with the suppliers plays an important role for PChome can offer to meet customers‘ satisfaction. The result revealed that the ability to achieve 24-hour delivery is steady at 0.88 in the long run. It is recommended that the approach outlined in this study be replicated in other e-shopping or third party companies.
6.2. Suggestions
The SM and FCM were chosen as the design methodologies is because they can be easily interpreted, since they clearly show the relationships between the different concepts and, at the same time, it is relatively easy and flexible to add or remove concepts, whenever necessary. Another problem that often arises in data gathering has to do with the fact that are often based on a survey, that is, the data are gathered through questionnaires, interviews, and so forth. In reality, there are several functions could be used to transform the value of the data.
A questionnaire is under development, which will be sent to expert specialists along with the description of the current 24-hour delivery model for future improvement of the model.
The assumption that the simulation is setting to be 5 workdays per run is a controversial one under practical consideration during this research; this means that PChome needs about 5 workdays to respond to the changes in the research design, which limits our interpretations. In the meantime, the simulation could be more complicated to discuss and explore even more issues. We set β =1 represents the weight don‘t affected by other concepts. In the future, maybe we can set every concept for different weight. We are hopeful that future research will be designed with much more sophistications allowing the ability to differentiate different point of view.
57
Reference
1. Andreou A.S., Mateou N.H., and Zombanakis G.A. (2005), Soft Computing for Crisis Management and Political Decision-making: the Use of Genetically Evolved Fuzzy Cognitive Maps, Soft Computing, pp. 194–210, 2005.
2. Axelrod, R. M. (1976), Structure of Decision: The Cognitive Maps of Political Elites, Princeton University Press.
3. Bart Kosko (1986), ―Fuzzy Cognitive Map‖, International Journal of Man-Machine Studies, Vol. 24, pp. 65-75.
4. Bellman, S., Lohse, G.L., Johnson, E.J. (1999), Predictors of online buying behavior, Communications of the Association for Computing Machinery,Vol.42, pp. 32–38.
5. Chan and Huang. (2004), ―A systems approach for the development of a sustainable community—the application of the sensitivity model (SM)‖, Journal of Environmental Management, Vol.72, pp.133–147.
6. Dickerson, J. A. and B. Kosko, Virtual Worlds as Fuzzy Cognitive Maps, 1993.
7. Dubois, D. and H. Prade, Fuzzy Sets and Systems: Theory and Applications, Academic Press, New York, 1980.
8. Eden and Ackermann. (2004), ―Cognitive Mapping Expert Views for Policy Analysis in the Public Sector‖, European Journal of Operational Research, Vol.152, pp.615-630.
9. Feng C.M., and Huang Y.K. (2007), ―Modeling the Determinants of Logistics Service Quality on Retailing Delivery Service for Online Shopping‖, Wireless Communications, Networking and Mobile Computing.
10. Feng C.M., and Huang Y.K. (2009), ―A Catastrophe Model for Developing Loyalty Strategies: a Case Study on Choice Behaviour of Pick-up Point for Online Shopping‖, International Journal of Services Operations and Informatics, Vol. 4, pp. 107 – 122.
11. FENG Yi-xiong, ZHENG Bing, &TAN Jian-rong. (2007), ―Exploratory study of logistics service quality scale based on online shopping malls.‖, Journal of Zhejiang University SCIENCE A, Vol.6, pp.926-931.
12. Farag, Sendy, Tim, Martin, & Jan. (2007), ―Shopping online and/or in-store? A structural equation model of the relationships between e-shopping and in-store shopping‖, Transportation Research Part A, Vol.41, pp.125-141.
13. J. T. Mentzer, D. j. Flint and G. T. M. Hult. (2001), Logistics service quality as a segment-customized process, J. of Mar., U. S. A., Vol. 65, pp. 82-104.
14. Jharkharia S., and Shankar R. (2007), Selection of Logistics Service Provider: An Analytic Network Process (ANP) Approach., Omega, Vol. 35, pp. 274-289.
58
15. Kang, S. Lee and J. Choi. (2004), Using Fuzzy Cognitive Map for the Relationship Management in Airline Service, Expert Systems with Applications, Vol. 26, pp.
545–555.
16. Lee, K. C. and H. S. Kim. (1997), ―A Fuzzy Cognitive Map-Based Bi-Directional Inference Mechanism: An Application to Stock Investment Analysis‖, Intelligence Systems in Accounting, Finance and Management, Vol. 6, pp. 41-57.
17. Lee, K. C., J. S. Kim, et al. (2002), Fuzzy Cognitive Map Approach to Web-mining Inference Amplification, Expert Systems With Applications, Vol. 22, pp. 197-211, . 18. Luis Rodriguez-Repiso, Rossitza Setchi, and Jose L. Salmeron. (2007), Modelling, IT
Projects Success with Fuzzy Cognitive Maps, Expert Systems with Applications, Vol.
32, pp. 543-559.
19. L.Y. Lin, Y.K. Huang and C.M Feng. (2009), Modeling Operation Dynamics of Third-party Logistics Providers with Fuzzy Cognitive Maps, National Chiao Tung University, Degree of Master of Engineering in Traffic and Transportation.
20. M. Hess. (2002), Shipping news: the implications of electronic commerce for logistics
20. M. Hess. (2002), Shipping news: the implications of electronic commerce for logistics