IRI-9529227 Janikow, Cezary U Missouri Saint Louis $35,544 - 12 mos REU and RUI: Fuzzy Decision Trees The objective of this research is to explore ways to utilize fuzzy representation to increase the representative power of symbolic decision trees. Decision trees are a popular and easilyunderstood inference mechanism for use in knowledge based systems. The use of fuzzy representations for knowledge has similar comprehensibility advantages, and is particularly natural in describing mixtures of continuous and discrete symbolic features. Fuzzy representations can advantageously be combined with decision trees. This project explores merging fuzzy representation and symbolic decision trees by modifying both the tree-building routines and the inference methods. The latter of these is quite challenging, and the results need to be empirically evaluated. The overall goal in doing this is to utilize the existing methodologies while preserving their important advantages, as mentioned above, in the combined forms.
