![]() | Fuzzy Inference | Mined Tree object – Field Usage tab | ![]() |
During inference the numeric/date fields behave as fuzzy objects. For information on fuzzy objects see Defining a Fuzzy Object in XpertRule. A MinedTree object allows two way splits on numeric/date fields in trees therefore during inference the S and Z membership functions for each field in the tree are created. The Reflex points, for Tree Fuzzification factor = 1, being set at:
Less than branch:
Numeric split value – average for field of all the records in this branch
Greater than or equal to branch
Numeric split value + average for field of all the records in this branch
The greater the Tree Fuzzification factor then the nearer the Reflex points are to the numeric split value.
Fuzzy inference usually produces more accurate results that crisp inference and the accuracy can be improved by experimenting with different Tree Fuzzification factors.
Mined Tree object – Field Usage tab