README for using randomized trees
                   =================================


This README is specifically for the code in the ExamplesRandomizedTrees subdirectory
of the installation directory.

See the README in the Examples subdirectory of the installation directory for the
main examples for using the DecisionTree module.

You can solve two different kinds of problems through the RandomizedTreesForBigData
class:

--- The problems that involve training a decision tree classifier with a dataset that
    suffers from serious imbalances in the populations available for the different
    classes.  If what you are facing is a two-class problem, you can directly use the
    RandomizedTreesForBigData class as currently programmed.  This class will create
    multiple decision trees for you, each trained with a mixture of training samples
    that consists of all the minority class samples and randomly drawn majority class
    samples.  The classification decisions for new data samples are based on majority
    voting by the individual decision trees.

--- If you are faced with a "big data" problem, in the sense that you have access to
    a very large training database of samples, you can use the
    RandomizedTreesForBigData class to create multiple decision trees by drawing
    training datasets randomly from your database.  Subsequently, the final
    classification decisions can be made by majority voting by the individual trees.

This directory contains the following scripts that you can use to become more
familiar with the RandomizedTreesForBigData class:

(1)   randomized_trees_for_classifying_one_test_sample_1.pl

      This illustrates the "looking for a needle in a haystack" function of the
      ExamplesRandomizedTrees class.
      
      
(2)   randomized_trees_for_classifying_one_test_sample_2.pl

      This illustrates how you may address a big data problem by constructing a set
      of decision trees, each based on random draw from your large training database.


(3)   classify_database_records.pl

      This illustrates evaluating the quality of the ensemble of decision trees
      constructed by ExamplesRandomizedTrees by aggregating all of the classification
      information for a reasonable large number of samples drawn randomly from a
      database.  See the comment block at the beginning of this script for what sort
      of diagnostic information the script puts out.