DecisionTree.h

Go to the documentation of this file.

#ifndef GRT_DECISION_TREE_HEADER
#define GRT_DECISION_TREE_HEADER

#include "../../CoreModules/Classifier.h"
#include "../../CoreAlgorithms/Tree/Tree.h"
#include "DecisionTreeNode.h"
#include "DecisionTreeThresholdNode.h"
#include "DecisionTreeClusterNode.h"
#include "DecisionTreeTripleFeatureNode.h"

GRT_BEGIN_NAMESPACE

class GRT_API DecisionTree : public Classifier
{
public:
    DecisionTree(const DecisionTreeNode &decisionTreeNode = DecisionTreeClusterNode(),const UINT minNumSamplesPerNode=5,const UINT maxDepth=10,const bool removeFeaturesAtEachSplit = false,const Tree::TrainingMode trainingMode = Tree::TrainingMode::BEST_ITERATIVE_SPILT,const UINT numSplittingSteps=100,const bool useScaling=false );
    
    DecisionTree(const DecisionTree &rhs);
    
    virtual ~DecisionTree(void);
    
    DecisionTree &operator=(const DecisionTree &rhs);
    
    virtual bool deepCopyFrom(const Classifier *classifier) override;
    
    virtual bool train_(ClassificationData &trainingData) override;
    
    virtual bool predict_(VectorFloat &inputVector) override;
    
    virtual bool clear() override;
    
    virtual bool recomputeNullRejectionThresholds() override;
    
    virtual bool save( std::fstream &file ) const override;
    
    virtual bool load( std::fstream &file ) override;
    
    virtual bool getModel( std::ostream &stream ) const override;
    
    DecisionTreeNode* deepCopyTree() const;

    DecisionTreeNode* deepCopyDecisionTreeNode() const;
    
    const DecisionTreeNode* getTree() const;

    Tree::TrainingMode getTrainingMode() const;
    
    UINT getNumSplittingSteps() const;
    
    UINT getMinNumSamplesPerNode() const;
    
    UINT getMaxDepth() const;
    
    UINT getPredictedNodeID() const;
    
    bool getRemoveFeaturesAtEachSplit() const;
    
    bool setTrainingMode(const Tree::TrainingMode trainingMode);
    
    bool setNumSplittingSteps(const UINT numSplittingSteps);
    
    bool setMinNumSamplesPerNode(const UINT minNumSamplesPerNode);
    
    bool setMaxDepth(const UINT maxDepth);
    
    bool setRemoveFeaturesAtEachSplit(const bool removeFeaturesAtEachSplit);

    GRT_DEPRECATED_MSG( "setRemoveFeaturesAtEachSpilt(const bool removeFeaturesAtEachSpilt) is deprecated, use setRemoveFeaturesAtEachSplit(const bool removeFeaturesAtEachSplit) instead", bool setRemoveFeaturesAtEachSpilt(const bool removeFeaturesAtEachSpilt) );

    bool setDecisionTreeNode( const DecisionTreeNode &node );

    static std::string getId();
    
    //Tell the compiler we are using the base class train method to stop hidden virtual function warnings
    using MLBase::save;
    using MLBase::load;
    using MLBase::train_;
    using MLBase::predict_;
    using MLBase::print;
    
protected:
    bool loadLegacyModelFromFile_v1( std::fstream &file );
    bool loadLegacyModelFromFile_v2( std::fstream &file );
    bool loadLegacyModelFromFile_v3( std::fstream &file );
    
    bool trainTree( ClassificationData trainingData, const ClassificationData &trainingDataCopy, const ClassificationData &validationData, Vector< UINT > features );
    DecisionTreeNode* buildTree(ClassificationData &trainingData, DecisionTreeNode *parent, Vector< UINT > features, const Vector< UINT > &classLabels, UINT nodeID );
    Float getNodeDistance( const VectorFloat &x, const UINT nodeID );
    Float getNodeDistance( const VectorFloat &x, const VectorFloat &y );
    
    DecisionTreeNode* decisionTreeNode;
    std::map< UINT, VectorFloat > nodeClusters;
    VectorFloat classClusterMean;
    VectorFloat classClusterStdDev;

    DecisionTreeNode *tree; 
    UINT minNumSamplesPerNode;
    UINT maxDepth;
    UINT numSplittingSteps;
    bool removeFeaturesAtEachSplit;
    Tree::TrainingMode trainingMode;

private:
    static RegisterClassifierModule< DecisionTree > registerModule;
    static const std::string id;
};

GRT_END_NAMESPACE

#endif //GRT_DECISION_TREE_HEADER