forest.cpp

   1 //
   2 //  forest.cpp
   3 //  Mothur
   4 //
   5 //  Created by Kathryn Iverson on 10/26/12.
   6 //  Copyright (c) 2012 Schloss Lab. All rights reserved.
   7 //
   8
   9 #include "forest.h"
  10
  11 /***********************************************************************/
  12 Forest::Forest(const std::vector < std::vector<int> > dataSet,
  13                const int numDecisionTrees,
  14                const string treeSplitCriterion = "gainratio",
  15                const bool doPruning = false,
  16                const float pruneAggressiveness = 0.9,
  17                const bool discardHighErrorTrees = true,
  18                const float highErrorTreeDiscardThreshold = 0.4,
  19                const string optimumFeatureSubsetSelectionCriteria = "log2",
  20                const float featureStandardDeviationThreshold = 0.0)
  21       : dataSet(dataSet),
  22         numDecisionTrees(numDecisionTrees),
  23         numSamples((int)dataSet.size()),
  24         numFeatures((int)(dataSet[0].size() - 1)),
  25         globalVariableImportanceList(numFeatures, 0),
  26         treeSplitCriterion(treeSplitCriterion),
  27         doPruning(doPruning),
  28         pruneAggressiveness(pruneAggressiveness),
  29         discardHighErrorTrees(discardHighErrorTrees),
  30         highErrorTreeDiscardThreshold(highErrorTreeDiscardThreshold),
  31         optimumFeatureSubsetSelectionCriteria(optimumFeatureSubsetSelectionCriteria),
  32         featureStandardDeviationThreshold(featureStandardDeviationThreshold)
  33         {
  34
  35     m = MothurOut::getInstance();
  36     globalDiscardedFeatureIndices = getGlobalDiscardedFeatureIndices();
  37     // TODO: double check if the implemenatation of 'globalOutOfBagEstimates' is correct
  38 }
  39
  40 /***********************************************************************/
  41
  42 vector<int> Forest::getGlobalDiscardedFeatureIndices() {
  43     try {
  44         //vector<int> globalDiscardedFeatureIndices;
  45         //globalDiscardedFeatureIndices.push_back(1);
  46
  47         // calculate feature vectors
  48         vector< vector<int> > featureVectors(numFeatures, vector<int>(numSamples, 0) );
  49         for (int i = 0; i < numSamples; i++) {
  50             if (m->control_pressed) { return globalDiscardedFeatureIndices; }
  51             for (int j = 0; j < numFeatures; j++) { featureVectors[j][i] = dataSet[i][j]; }
  52         }
  53
  54         for (int i = 0; i < featureVectors.size(); i++) {
  55             if (m->control_pressed) { return globalDiscardedFeatureIndices; }
  56             double standardDeviation = m->getStandardDeviation(featureVectors[i]);
  57             if (standardDeviation <= featureStandardDeviationThreshold){ globalDiscardedFeatureIndices.push_back(i); }
  58         }
  59
  60         if (m->debug) {
  61             m->mothurOut("number of global discarded features:  " + toString(globalDiscardedFeatureIndices.size())+ "\n");
  62             m->mothurOut("total features: " + toString(featureVectors.size())+ "\n");
  63         }
  64
  65         return globalDiscardedFeatureIndices;
  66     }
  67         catch(exception& e) {
  68                 m->errorOut(e, "Forest", "getGlobalDiscardedFeatureIndices");
  69                 exit(1);
  70         }
  71 }
  72
  73 /***********************************************************************/
  74