Mining Matrix Pattern from Mobile Users - Distributed Artificial Intelligence, Agent Technology, and Collaborative Applications - page 239

logical event must be related to the domain, also known as theme of that layer. For instance, the single
layer matrix in Figure 2 shows the frequency of occurrence of designer clothing. The designer clothing
serves as the domain for the layer of matrix layer. A confidence threshold is determined, depending on
the size of the matrix and the size of the population in order to determine which values can be considered
relevant for problem solving and which are not. The confidence threshold determines the sensitivity of
the result of mobile user data mining. For instance, the confidence threshold for the matrix in Figure 2
is 40%, and therefore only static nodes with logical event confidence occurrence for designer clothing
of equal or greater than 40% ( X axis = 7; Y axis = 8) are taken into consideration.
A window size, which is a specified timeframe for which the source data from mobile users are
gathered, is also required in order to define the scope of the time series that the decision maker wishes
to take into consideration. The data collected within the window size timeframe limits the amount of
source data to be gathered and this could be based on particular time of the day and can be adjusted by
the data miner. By having a window size, it serves as the base number for the calculation of confidence,
as confidence = frequency / window size . The result of this process is a presentation of all physical loca-
tions that contains high activity of the particular logical event.
Figure 3 illustrates the result for single layer matrix pattern. There are three relationships. Relation-
ships are connected from one unit to another. The list of relationships are: (0,0 - 8,0) = 85, (0,0 - 1,8)
= 80, (8,0 - 1,8) = 75. It can be seen that low confidence values are all removed and the relationship in
between each static nodes are calculated and sorted according to highest confidence first. It shows that
there is a significant relationship between these static nodes.
The following steps represent the important phases in the process of generating a single layer matrix
pattern. It is important that these steps are followed closely in the sequential order in order to preserve the
overall meaning of single layer matrix. The outcome of this algorithm is the single layer matrix, which
shows the relationship in between different static nodes with reference to single logical domain.
Step1:DeineSingleLayerMatrix
Figure 4 shows the first step in the single layer matrix based mobile user data mining, which involves the
definition of the single layer matrix. A single layer matrix is a two-dimensional matrix within the same
level of plane. This single layer matrix provides the physical mapping of the positions of static nodes
in the mobile environment. The size of the single layer matrix can be defined by using two variables, x
and y, that represent the horizontal axis and vertical axis respectively.
It is important to note that the horizontal length and the vertical length need not to be the same, as
long as the single layer matrix covers the geographical area in which the decision maker is interested. The
definition of single layer matrix must be as one single matrix, that is, when two locations are required,
they have to be defined within the same matrix in order to achieve the effect of comparisons.
Figure 4. Algorithm for defining single layer matrix
Matrix Create_Matrix(Single Layer, Hsize, Vsize) {
Array = Create_Array(Hsize, Vsize);
Initialize_Array(0);
Return *Array;
}