******************************************************************************** * * * Copyright J.R. Quinlan, 1989, 1990, 1991, 1992. This software may not be * * distributed without permission of the copyright holder. * * * ******************************************************************************** The C4.5 induction system ------------------------- The C4.5 system consists of four principal programs: 1) the decision tree generator ('c4.5'), 2) the production rule generator ('c4.5rules'), 3) the decision tree interpreter ('consult'), and 4) the production rule interpreter ('consultr'). This release also contains stuff that may be helpful in running experiments: 5) a csh shell script for cross-validation ('xval.sh'), and 6) two ancillary programs ('xval-prep' and 'average'). There is a Unix-style (brief) manual entry for the main programs and for their verbose output mode. A better guide to using the system is provided in my book, 'C4.5: Programs for Machine Learning', Morgan Kaufmann 1992. [Note: changes since the book was published are recorded in the file Modifications in the Src directory.] Installation ------------ The system has been targeted to Berkeley BSD4.3. Some other environments (e.g. SystemV) may require the use of additional libraries etc; if you have any trouble compiling the system, look at the comments in the Makefile. In particular, if you are running in an environment other than Berkeley BSD4.3 you may need to search additional libraries, e.g. for the random number generator 'random'. This requires the addition of loader instructions of the form -l* after '-lm'; you might try '-lbsd'; if this doesn't work, consult your local Unix guru. To create the executable binary files, go into the 'Src' directory by typing: cd Src and type: make all This will create the binary files 'c4.5', 'c4.5rules', 'consult', 'consultr', 'xval-prep' and 'average'. These, together with the shell script, can be moved into your installation's directory for local software, e.g. '/usr/local/bin' by typing: mv c4.5 c4.5rules consult consultr xval-prep average xval.sh /usr/local/bin The object files created during compilation may be removed to save space by typing: rm *.o In the 'Doc' directory, there is a manual entry file for each of the programs, as well as the files 'verbose.1' and 'verbrules.1' which explain the verbose output produced by 'c4.5' and 'c4.5rules' respectively. These can be installed in your local manual by moving them to your installation's manual directory for volume 1, usually '/usr/man/man1' by typing, for instance: cd ../Doc mv *.1 /usr/man/man1 (You may need superuser privileges to put files in this directory.) Examples -------- The directory 'Data' contains some sample datasets. Many interesting data collections are maintained by the University of California, Irvine, in a data repository. For more information on how to access files from this repository, send email to ml-repository@ics.uci.edu Problems -------- If you discover any bugs in the system, please notify me by email to quinlan@cs.su.oz.au or by writing to me: J.R. Quinlan Basser Department of Computer Science Madsen Building F09 University of Sydney Sydney Australia 2006