next up previous contents
Next: Why Differences Occur Up: Data Mining and Influence Previous: The Bayesian Network

Differences in Results

Generally, the retrieved case numbers are higher for the CBRDM system than for KATE. The reason for this is that if there are more than one case that matches the ``new'' case with the same similarity, CBRDM returns the last match, while KATE returns the first.

If we look at the series with results given in chapter 5, we see that the output cases from the two methods are only similar for query 4 of series 4 (Tab 5.6). The feature values are the same, but the case numbers are different. In Fig 5.4 the number of differences in feature values between the cases retrieved with the two methods are shown. 43 of the pairs of cases retrieved differ by three or more feature values. On average, every case differs by 4.5 feature values of a total of 9. So, it seems that the feature values retrieved by the systems are indeed different.



Torgeir Dingsoyr
2/26/1998