Gregory Piatetsky-Shapiro answers:
The best description is in R. Quinlan book
C4.5: programs for machine learning, ISBN:1-55860-238-0,
Morgan Kaufmann, 1993.
See also
Ross Quinlan personal page which has many of his papers, and
a decision-tree module of my Data Mining Course,
at www dot kdnuggets dot com /data_mining_course/dm7-dec-tree-c45.ppt
gioseza adds a warning:
in Quinlan's own words, the pruning strategy used by C4.5 is a "heuristic with questionable underpinnings."
gioseza suggests to study the more statistically grounded approach used by CART.
Note: ID3 (id3) algorithm is a predecessor of C45 algorithm.
|