Result and Discussion

GC N DG

V DG

  . (3.17)

Rather than locating decision maker into a certain group by measuring degree of interaction, in this benchmark method, the group with highest GCM was selected as decision group. For example, consider a social network with two groups (see Figure 3.7). Both group A and B have 4 members, and the entire social network has 15 members. 2 out of 11 members who are not in group A are connected with the members in group A, and 3 of them are connected with group B. Therefore, group B has higher group centrality than A.

Figure 3.7 Example of group centrality

3.5 Result and Discussion

In this experiment every alternative and store presented to decision makers were collected, and the average product usefulness level of different methods and groups is plotted in Figure 3.8. As shown, the proposed mechanism attracted decision makers to be more satisfied on the products and stores than other methods. Moreover, as shown in Figure 3.9, the average usefulness level of stores suggested by proposed mechanism was also higher than other methods. To further examine if there were significant differences in average usefulness level for products and stores, a statistical method was required.

-30-

Figure 3.8 Average usefulness level about product ranking

Figure 3.9 Average usefulness level about store ranking

-31-

Two-way analysis of variance (ANOVA) is a statistical analysis in which two independent factors are examined with regard to their impact on a dependent variable and on one another. To test the impact of method used and user group on average product and store usefulness level, in this work two-way ANOVA was used. As shown in Table 3.7, the method used in the experiment has impact on the average product satisfaction as the test result is significant at 0.05 (as 0.00<0.05). In contract, the user group has no impact as 0.92>0.05. For the same reason, based on Table 3.8 the store satisfaction can only be influenced by method used during the experiment.

Table 3.7 Tests of between-subjects effects for average product usefulness level

Dependent Variable: Average Product Usefulness Level SOURCE TYPE III SUM OF

Table 3.8 Tests of between-subjects effects for average store usefulness level

Dependent Variable: Average Store Usefulness Level SOURCE TYPE III SUM OF

Post hoc tests such as Tukey's test most commonly compare every group mean with every other group mean. Knowing that the methods used in the experiments could affect stay time and usefulness level, Tukey’s test was used to see if there is a significant difference between different methods. As observed from Table 3.9 and Table 3.10, there were significant differences between the proposed mechanism and other benchmark methods; also the average product and store usefulness level were higher than other methods. Based on these statistic results, it is likely that proposed approach is more effective when compared with other methods.

-32-

Table 3.9 Multiple comparisons of product voting usefulness level

(I) METHOD (J) METHOD MEAN DIFFERENCE (I-J)

Random

*. The mean difference is significant at the .05 level.

Table 3.10 Multiple comparisons of store voting usefulness level

(I) METHOD (J) METHOD MEAN DIFFERENCE (I-J)

*. The mean difference is significant at the .05 level.

Besides the stay time and usefulness survey, in the experiment the decision makers were also asked to provide ranking of the product candidates list so that the ranking result from social support mechanism can be compared. Kendall's  is a measure of correlation, and so measures the similarity of the ranking between two lists r and _a r . It is a coefficient that represents the degree of concordance between two columns b

-33-

of ranked data. It requires that the two variables are paired observations, for example, ranking from teacher and student for each book in the sample. Then, provided both variables are at least ordinal, it would be possible to calculate the correlation between them. For each variable separately the values are put in order and numbered, 1 for the lowest value, 2 for the next lowest and so on. Kendall's tau takes values between -1 and +1. The greater the numbers of inversions, the smaller the coefficient will be. A positive correlation indicating that the ranks of both observations increase together whilst a negative correlation indicates that as the rank of one variable increases the other one decreases. Kendall's tau is formulated as [41]:

( , )_{a b} ^{C D},

Where N and _C N are the number of concordant and discordant pair. An example for _D calculating Kendall's  is given in Table 3.11, and the value of Kendall's  in this similarity of the ranking result from system and users.

Table 3.11 Example of Kendall’s Tau value calculating

Teacher Student N_C N_D

Table 3.12 summarizes the Kendall’s  values. As 70% (75/108) of these values are positive, it is likely to conclude that the proposed mechanism can provide good enough ranking information for users.

-34-

Table 3.12 Kendall’s  value of system and user ranking similarity

Group Member Kendall's  Value

No. of positive value: 75, No. of negative value: 33(21 from store, 12 from product)

3.6 Chapter Summary

In this chapter, social network analysis, social influence and adaptive majority voting were used to design a social support mechanism for product purchasing decision-making process. From the viewpoint of academic contribution, by utilizing MRQAP analysis the relations between friendship and social interactions in online social network sites were tested and verified. Social network analysis skills were used to profile individual users within online social networks. A home group locating method was also proposed to place the decision maker in right group so that the members can provide better suggestion. Furthermore, an adaptive voting mechanism was also suggested to further improve the majority voting on social network related research. An empirical study further proved the feasibility and effectiveness. This research successfully introduced the decision process theory and social psychology into the development of social network-based application. Besides, this study also extended

-35-

the concept of decision support system development to utilize social network platform.

From the viewpoint of practice, this study showed a feasible way to develop a social network-based decision support system together with the related techniques for the purpose of product purchasing decision-making. By dividing the system framework into modules, those who are interested in developing such kind of applications can further improve the system by plugging in new modules as needed.

36