Determining the incremental worth of members of an aggregate set through difference-based induction
Abbreviated Journal Title
Int. J. Intell. Syst.
Computer Science, Artificial Intelligence
Calculating the incremental worth or weight of individual components of an aggregate set when only the whole set's total worth or weight is known is a problem common to several domains. Here we describe an algorithm that induces such incremental worth from a database of similar (not identical) aggregate sets. The algorithm focuses on finding aggregate sets in the database exhibiting minimal differences in corresponding components (attributes and values). This procedure isolates dissimilarities between nearly similar aggregate sets so any difference in worth between sets is attributed to them. The algorithm builds a classification tree similar to those of ID3 and C4.5 distributes all aggregate sets in the database according to their attributes and values; and groups together those with the same attributes and values. Each leaf of the classification tree then contains a group of aggregate sets identical to each other insofar as their attributes and values. Groups' members belonging to two sibling leaves (having the same immediate parent) differ from each other in the value of exactly one attribute. Thus, any difference in the worth of sets in those groups can be attributed to that difference. The worth of the aggregate sets in these groups can be averaged when data are noisy. This algorithm works well when applied to real-estate appraisal domain. (C) 1999 John Wiley & Sons, Inc.
International Journal of Intelligent Systems
"Determining the incremental worth of members of an aggregate set through difference-based induction" (1999). Faculty Bibliography 1990s. 2649.