next up previous contents
Next: Integrating DM and CBR Up: Data Mining Previous: Missing values

An Example

To give an example of possible use of Data Mining, let us return to the example given in chapter 2. Let us say that we have a database of accidents on a platform which could have the following format:


ACTIVITY       DEVIATIONS                     INCIDENT       INJURY
=========================================================================
CARRY_PIPE     LEVER & OILSPILL & NO_GLOVES   OP_SLIPPED     HAND_CRUSHED
CARRY_PIPE     OILSPILL                       OP_SLIPPED     FOOT_CRUSHED
...

Where the activity when the injury occurred is described, deviations from standard procedures, which incident that caused the injury and the injury itself if described. We can then try to make a clustering of the different injury types to learn more about the injuries, and then find measures to reduce them. An obvious reaction after the two accidents in the database above would be to make sure that oil is not spilt on the platform.


  
Figure 3.7: A Bayesian Network for Accident Causes.
\begin{figure}

\begin{center}

\scalebox {0.4}{\includegraphics*[04cm,09cm][26cm,28cm]{oilbay.eps}}

\end{center}\end{figure}

We could also try to make a network of events which can be said to cause the injury, and for instance compute probabilities of each event and give a certain cost for correction so that we are able to avoid similar accidents cost-effectively. Such a network is given in Fig 3.7.



Torgeir Dingsoyr
2/26/1998