view release on metacpan or  search on metacpan

ExamplesBoosting/boosting_for_bulk_classification.pl  view on Meta::CPAN

#!/usr/bin/env perl

### boosting_for_bulk_classificaiton.pl

##  Call syntax example:

##    boosting_for_bulk_classification.pl  training6.csv  test6.csv   out6.csv

##  This script demonstrates how a boosted decision-tree classifier can be used
##  to carry out bulk classification of all your test samples in a file.

use strict;
use warnings;
use Algorithm::BoostedDecisionTree;

die "This script must be called with exactly three command-line arguments:\n" .
    "     1st arg: name of the training datafile\n" .
    "     2nd arg: name of the test data file\n" .     
    "     3rd arg: the name of the output file to which class labels will be written\n" 
    unless @ARGV == 3;

my $debug = 0;

my ($training_datafile, $test_datafile, $outputfile) = @ARGV;

my $training_file_class_name_in_column       = 1;
my $training_file_columns_for_feature_values = [2,3];
my $how_many_stages                          = 4;

my (@all_class_names, @feature_names, %class_for_sample_hash, %feature_values_for_samples_hash,
    %features_and_values_hash, %features_and_unique_values_hash,
    %numeric_features_valuerange_hash, %feature_values_how_many_uniques_hash);

my $boosted = Algorithm::BoostedDecisionTree->new(
                              training_datafile => $training_datafile,
                              csv_class_column_index => $training_file_class_name_in_column,
                              csv_columns_for_features => $training_file_columns_for_feature_values,
                              entropy_threshold => 0.01,
                              max_depth_desired => 8,
                              symbolic_to_numeric_cardinality_threshold => 10,
                              how_many_stages => $how_many_stages,
                              csv_cleanup_needed => 1,
              );

print "Reading and processing training data...\n";
$boosted->get_training_data_for_base_tree();

##   UNCOMMENT THE FOLLOWING STATEMENT if you want to see the training data used for
##   just the base tree:
$boosted->show_training_data_for_base_tree();

# This is a required call:
print "Calculating first-order probabilities...\n";
$boosted->calculate_first_order_probabilities_and_class_priors();

# This is a required call:
print "Constructing base decision tree...\n";
$boosted->construct_base_decision_tree();

#   UNCOMMENT THE FOLLOWING TWO STATEMENTS if you would like to see the base decision
#   tree displayed in your terminal window:
#print "\n\nThe Decision Tree:\n\n";
#$boosted->display_base_decision_tree();

# This is a required call:
print "Constructing the rest of the decision trees....\n";
$boosted->construct_cascade_of_trees();

#  UNCOMMENT the following statement if you wish to see the class labels for the
#  samples misclassified by any particular stage.  The integer argument in the call
#  you see below is the stage index.  Whe set to 0, that means the base classifier.
$boosted->show_class_labels_for_misclassified_samples_in_stage(0);


##  UNCOMMENT the next statement if you want to see the decision trees constructed
##  for each stage of the cascade:
print "\nDisplaying the decision trees for all stages:\n\n";
$boosted->display_decision_trees_for_different_stages();

### NOW YOU ARE READY TO CLASSIFY THE FILE-BASED TEST DATA:
get_test_data_from_csv();

open FILEOUT, ">$outputfile"
    or die "Unable to open file $outputfile for writing out classification results: $!";

my $class_names = join ",", sort @{$boosted->get_all_class_names()};

my $output_string = "sample_index,$class_names\n";

print FILEOUT $output_string;

foreach my $item (sort {sample_index($a) <=> sample_index($b)} keys %feature_values_for_samples_hash) {
    my $test_sample =  $feature_values_for_samples_hash{$item};
    $boosted->classify_with_boosting($test_sample);
    my $classification = $boosted->trust_weighted_majority_vote_classifier();
    my $output_string = sample_index($item);
#    $output_string .=  "," + $classification[11:]    
    $output_string .=  ",$classification";
    print FILEOUT "$output_string\n";
}

print "Majority vote classifications using boosting written out to $outputfile\n";

############################  Utility Routines #################################