Information Technology Reference
In-Depth Information
the current time (2011-03-11T17:00:00), and the current location (Chiyoda-ku) of the
active user. The result of this query is shown in Figure 9. Therefore, we can say that our
action network is working properly with RDF queries.
5.3
Missing Activity Prediction
To evaluate our proposed approach, we first created correct action data of 3,900 Twitter
users in Tokyo, after the massive earthquake occurred. Secondly, we repeated 10 times
of the following experiment.
Ta b l e 2 .
Recall of Deleted Activity Data
Method
Action
Location Action and Location
Baseline@
31.48% 43.09%
27.56%
Our method 69.23% 76.92%
43.59%
1. Randomly select 39 users as the active users.
2. Randomly delete activity data of these active users.
3. Consider the active users' names and time of deleted activities as input data, using
our approach to determine whether the deleted activity data is reproduced or not.
The average results are shown in Table 2. In this table, baseline is the following method:
1. Look up the most similar user
most similar
u
a
to the active user
u
a
.
2. Based on
most similar
u
a
, we predict missing activity of
u
a
.
From the above results, we can say that:
-
Our approach can reproduce 69.23% of missing actions, 76.92% of missing loca-
tions, and 43.59% of missing activities (both of action and location).
-
Not only the most similar user, our method considers all similar users and candidate
actions. This is the reason why our method outperforms baseline.
5.4
Application of Timeline Action Network
If data on Twitter is real-time data, then we can say that TiAN reflects real-world activ-
ities in real-time. By using SPARQL (SPARQL Protocol and RDF Query Language),
computers can grasp situations of trains, evacuation centers, food shops, etc. Therefore,
we can use TiAN to find the nearest available evaluation center for disaster victims.
The computers also can recommend “what should to do” for the active user, based
on action patterns of other users in TiAN. For example, as shown in Figure 10, based
on past actions of other users, the computer can recommend
{
act 05, act 06, act 08
}
or
{
act 05, act 06, act 07, act 08
}
for the active user at Shinjuku station.
