Information Technology Reference
In-Depth Information
unnoticeable (e.g. in the winter of 1997/98), while in other years doctors and
pharmacists even run out of vaccine (e.g. in the winter of 1995/96). Furthermore,
figure 1 shows that influenza waves occurred in February and March. However, we
know that this is probably accidental and in Germany a wave may already start much
earlier (the last influenza epidemic started in December 1995).
Influenza waves are difficult to predict, because they are cyclic, but not regular
[6]. Because of the irregular cyclic behaviour, it is insufficient to determine average
values based on former years and to give warnings as soon as such values are
noticeably overstepped. So, we have developed a method that combines Temporal
Abstraction [7] with Case-based Reasoning [8, 9]. The idea is to search for former,
similar courses and to make use of them for the decision whether early warning is
appropriate.
Viboud et al. [10] apply the method of analogues [11], which originally was
developed for weather forecasting. It also takes former, similar courses into account.
However, the continuations of the most similar former courses are used to predict
future values, e.g. the influenza incidences of next week. Instead, we intend to
discover threatening influenza waves in advance and to provide early warnings against
them.
4000
3500
3000
2500
1997/1998
1998/1999
1999/2000
2000/2001
2000
1500
1000
500
0
1
3
5
7
9
11
13
15
17
19
21
23
25
Weeks
Fig. 1.
Influenza courses for Mecklenburg-Western Pomerania from October till March. The
1
st
week corresponds to the 40
th
week of the calendar and 14
th
week to the 1
st
week of the next
year.
