ClusterTree.h

Go to the documentation of this file.

#ifndef GRT_CLUSTER_TREE_HEADER
#define GRT_CLUSTER_TREE_HEADER

#include "../../CoreModules/Clusterer.h"
#include "../../CoreAlgorithms/Tree/Tree.h"
#include "ClusterTreeNode.h"

GRT_BEGIN_NAMESPACE

class GRT_API ClusterTree : public Clusterer
{
public:
    ClusterTree(const UINT numSplittingSteps=100,const UINT minNumSamplesPerNode=5,const UINT maxDepth=10,const bool removeFeaturesAtEachSplit = false,const Tree::TrainingMode trainingMode = Tree::BEST_ITERATIVE_SPILT,const bool useScaling=false,const Float minRMSErrorPerNode = 0.01);
    
    ClusterTree(const ClusterTree &rhs);
    
     virtual ~ClusterTree(void);
    
    ClusterTree &operator=(const ClusterTree &rhs);

    virtual bool deepCopyFrom(const Clusterer *cluster) override;
    
    virtual bool train_(MatrixFloat &trainingData) override;
    
    virtual bool predict_(VectorFloat &inputVector) override;
    
    virtual bool clear() override;
    
    virtual bool print() const override;
    
    virtual bool saveModelToFile( std::fstream &file ) const override;
    
    virtual bool loadModelFromFile( std::fstream &file ) override;

    ClusterTreeNode* deepCopyTree() const;
    
    const ClusterTreeNode* getTree() const;
    
    UINT getPredictedClusterLabel() const;
    
    Float getMinRMSErrorPerNode() 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);

    bool setMinRMSErrorPerNode(const Float minRMSErrorPerNode);
    
    //Tell the compiler we are using the base class train method to stop hidden virtual function warnings
    using MLBase::saveModelToFile;
    using MLBase::loadModelFromFile;
    using MLBase::train;
    using MLBase::predict;

    static std::string getId();
    
protected:
    Node *tree; 
    UINT minNumSamplesPerNode;
    UINT maxDepth;
    UINT numSplittingSteps;
    bool removeFeaturesAtEachSplit;
    Tree::TrainingMode trainingMode;
    Float minRMSErrorPerNode;
    
    ClusterTreeNode* buildTree( const MatrixFloat &trainingData, ClusterTreeNode *parent, Vector< UINT > features, UINT &clusterLabel, UINT nodeID );
    bool computeBestSplit( const MatrixFloat &trainingData, const Vector< UINT > &features, UINT &featureIndex, Float &threshold, Float &minError );
    bool computeBestSplitBestIterativeSplit( const MatrixFloat &trainingData, const Vector< UINT > &features, UINT &featureIndex, Float &threshold, Float &minError );
    bool computeBestSplitBestRandomSplit( const MatrixFloat &trainingData, const Vector< UINT > &features, UINT &featureIndex, Float &threshold, Float &minError );

private:
    static RegisterClustererModule< ClusterTree > registerModule;
    static const std::string id;
};

GRT_END_NAMESPACE

#endif //GRT_CLUSTER_TREE_HEADER