Information Technology Reference
In-Depth Information
(0,0 - 1,8) = 77.5, (0,0 - 8,0) = 72.5. The combined relationship between fashion and leisure domain is:
(0,0 - 8,0) = (85 + 72.5) / 2 = 78.75%; (0,0 - 1,8) = (80 + 77.5) / 2 = 78.75%; (8,0 - 1,8) = (75 + 80) / 2
= 77.5%.
The inter-domain relationships are calculated based on the average between the two confidences
found in the same particular space in different layers. In some instances when the original single layer
confidence threshold is set too low, it is possible for the relationship confidence returned low as well.
Therefore, a relationship confidence threshold has to be set in order to limit the amount of significant
relationships to avoid pattern overloading. All relationships with the relationship confidence lesser than
the relationship threshold will be discarded and not taken into consideration.
The relationship confidence within the same layer will only help the decision maker in making a
decision which requires single factor related to the specific logical domain. Often, the case is that mul-
tiple factors need to be considered. For instance, when setting up a shop that sells designer clothing,
one must take into consideration factors such as the income level of the surrounding, the presence of
other fashion design domains in the area, the presence of any artistic domains in the area, and finally,
the presence of any potential customers in the area.
In order to take multiple factors into consideration, multiple layers of the matrix can be selected.
After the list of layers is selected, the multi-domain relationship confidence is calculated. The formula
for evaluating the multi-domain relationship confidence is:
Multi-Domain Relationship Confidence =
(Layer[1].Relationship_Confidence + Layer[2].Relationship_Confidence +
...
+ Layer[n].Relation-
ship_Confidence) / n
. It is the average of the combination of all relationships in the multiple layers.
Finally, a multi-domain relationship confidence threshold is set in order to adjust the sensitivity of
the pattern found. In order to achieve this, the multi-domain relationship confidence is compared against
the multi-domain relationship confidence threshold. For all multi-domain relationship confidence lesser
than the multi-domain relationship threshold, they are discarded to ensure the accuracy of the result.
The result of the multi-layer matrix provides a clear examination of the relationships between static
nodes, not just within the same layer, but also between multiple layers. By performing analysis into
multiple layers in which each layer represents a logical domain, this methodology provides an innova-
tive approach in finding out useful knowledge from mobile users and mobile stations which takes both
the physical and logical considerations into account at the same time.
The following steps represent the essential parts of the process of using multi-layer matrix in order
to perform mobile user data mining. It is important to note that in a multi-layer matrix, another dimen-
sion is added and therefore, three threshold values are required.
Step1:DeineMulti-LayerMatrix
The first step in this mobile user data mining exercise is to define the multi-layer matrix. A multi-layer
matrix contains three dimensions. It is essential to note that there is no need to have all three variables
the same size. Although they can be configured to be the same size, the decision rests on the decision
maker and the problem that needs to be solved. A multi-layer matrix contains multiple layers of single
layer matrix. The two-dimensional single layer matrix has to be first defined by defining the horizontal
and vertical length of the single layer matrix.
After the size of the single layer matrix, the number of layers in the multi-layer matrix is dependent
on how many independent logical domains are present in the mobile data-mining environment. The
user can decide this by looking at how many logical domains or themes are present and also how many
Search WWH ::
Custom Search