Information Technology Reference
In-Depth Information
can evaluate the provided recommend service and the provided information first.
Users are also allowed to provide description information of each spot using in-
formation tags. These tags will be included in the database and for future utiliza-
tion. In this experiment, the tag information in database is previously generated
from students who had visited the spots of each bicycle route. During the experi-
ment process, the information input function is temporary disconnected to main
database so as to make sure users will receive recommendations from the same
database. Nevertheless, their contributions are manually added into database for
further analysis. Again, the only difference in each experiment is the recommend
mechanism. We will examine and analysis differences between the three recom-
mend methods according to the five evaluate dimensions. Following are the
experiment results of each paired group comparison.
4.2 Experiment Result
We use the independent t-test to examine the difference between each paired
group. The decision quality comparisons include three pairs: (1) TF-IDF method
vs. TF method, (2) CTD method vs. CTD method, and (3) TF-IDF method vs.
CTD method.
When comparing the TF-IDF method with traditional TF method, there was a
significant difference between the two groups. This significant difference between
two groups was confirmed by independent t-test (p<0.05) in each dimensions. The
TF-IDF method provides high quality information for users than traditional rec-
ommend method. Users feel the recommend information using TF-IDF could pro-
vide high quality information to improve their decision. Unsurprisingly, using
term frequency (TF) could only bring the popular issues for user. However, the
provided information may be too conceptual that did not addressing user's need
for detailed entertainment information.
The comparing outcome of CTD method with TF method is similar to previous
experiment. The independent t-test (p<0.05) indicate the CTD recommendation
provide high quality information than the TF method in all five dimensions. We
can see the information using CTD is more addressing the category of the leisure
information than TF method.
However, when comparing the TF-IDF method with CTD method, there is
no significant difference between the two recommending mechanism in all
dimensions. We will explore the behavioral analysis in this issue in the following
subsection.
In summary, when compare with using pure term frequency for recommenda-
tion, both the TF-IDF and CTD methods are comparatively successful than tradi-
tional TF method. The proximal social intelligence brings latest information for
each leisure e-service users. Moreover, the provided information contains personal
opinions and suggestions that are not available from traditional information
sources. The i-Bike service provide a platform for users to collaborate with others
to filer and extract useful information that is focusing on user's urgent needs. The
satisfaction of information quality in each recommendatory mechanism is im-
proved than using TF method.
