A Density-Based Clustering of Spatio-Temporal Data - New Contributions in Information Systems and Technologies - page 44

Table 2. Summary of the Temporal Dimension for Elk Data and Deer Data

Elk Data Set Deer Data Set

For eps=59
MinLns=10

Number of
Segments

Avg. Seg.
Time

For eps=30
MinLns=6

Number of
Segments

Avg. Seg.
Time

Cluster 1 3036 7.4 hours Cluster 1 1551 5.93 hours

Cluster 2 26 5.75 hours Cluster 2 37 5.78 hours

Cluster 3 28 5.8 hours Cluster 3 15 5.33 hours

Cluster 4 34 6 hours Cluster 4 29 5.13 hours

Cluster 5 48 5.25 hours Cluster 5 15 5.45 hours

Cluster 6 19 5.14hours

5 Discussion

From the experimental results in this study, it can be concluded that the time dimension
and distance of travel are not in direct relation. This means that longer distances do not
necessarily mean longer time of travel. This is clearly shown in Table 1 for the
Hurricane data set experiment. Table 1 shows Cluster 1 with a density of 889 line
segments and average time of travel of 2.84 days. Looking at the length of Cluster 1 in
Figure 4, it is concluded that sub-trajectories of hurricanes that pass through this cluster
have very high speeds to cross such distance in such short time. It is also predicted that
this area might not have any obstacles that could slow down the motion of a passing
hurricane or deviate its direction of motion. On the other hand, from the experimental
results carried out for the Elk and Deer data sets, some conclusions are made. The
longer the cluster, the higher the density of line segments in it, the greater the average
segment time of travel. Table 2 represents Cluster 1 for Elk movement with a density
of 3036 line segments. Compared to the density of the other clusters for this run, this
number is very big. Cluster 1 has the greatest average segment time, 7.4 hours. It is
predicted that this great time is due to the high density of elks which slow each other
down therefore take more time. In addition that the cluster has the largest length. For
the deer results in Table 2 this prediction is supported. Cluster 1 for Deer movement is
the highest in density with 1551 sub-trajectories and the greatest average segment time,
5.93 hours. This case is a little bit more different from the Elk experiment. The average
segment time for these clusters in the Deer data set are all relatively close while the
distances of travel seen in Figure 6 are different in length. It is predicted that for
Cluster 1 there are no obstacles that slow down the motion of the deer, while in
Clusters 3, 5 and 6 there could be obstacles that slow down the motion, since the
distances of travel are short but the average segment time of travel is nearly as large as
those of Clusters 1 and 2.