gbt_cls_traversed_model_builder.cpp

/* file: gbt_cls_traversed_model_builder.cpp */
/*******************************************************************************
* Copyright 2014-2019 Intel Corporation.
*
* This software and the related documents are Intel copyrighted  materials,  and
* your use of  them is  governed by the  express license  under which  they were
* provided to you (License).  Unless the License provides otherwise, you may not
* use, modify, copy, publish, distribute,  disclose or transmit this software or
* the related documents without Intel's prior written permission.
*
* This software and the related documents  are provided as  is,  with no express
* or implied  warranties,  other  than those  that are  expressly stated  in the
* License.
*******************************************************************************/

/*
!  Content:
!    C++ example of of gradient boosted trees classification model
!    building from traversed gradient boosted trees classification model
!
!    The program trains the gradient boosted trees classification model, gets
!    pre-computed values from nodes of each tree using traversal and build
!    model of the gradient boosted trees classification via Model Builder and
!    computes classification for the test data.
!******************************************************************************/

#include "daal.h"
#include "service.h"

using namespace std;
using namespace daal;
using namespace daal::algorithms::gbt::classification;

const string trainDatasetFileName = "../data/batch/df_classification_train.csv";
const string testDatasetFileName  = "../data/batch/df_classification_test.csv";
const size_t categoricalFeaturesIndices[] = { 2 };
const size_t nFeatures  = 3;  /* Number of features in training and testing data sets */

/* Gradient boosted trees training parameters */
const size_t maxIterations = 50;
const size_t minObservationsInLeafNode = 8;

const size_t nClasses = 5;  /* Number of classes */
size_t nTrees = 0;
struct Node
{
    Node* left;
    Node* right;
    double response;
    size_t featureIndex;
    double featureValue;
    Node(double rs, size_t fi, double fv): left(NULL), right(NULL), response(rs), featureIndex(fi), featureValue(fv)
    {}
    Node(): left(NULL), right(NULL), response(0), featureIndex(0), featureValue(0)
    {}
};

struct Tree
{
    Node* root;
    size_t nNodes;
    ~Tree()
    {
        if(root)
        {
            delete root;
        }
    }
};

struct ParentPlace
{
    size_t parentId;
    size_t place;
    ParentPlace(size_t _parent, size_t _place): parentId(_parent), place(_place)
    {}
    ParentPlace(): parentId(0), place(0)
    {}
};

class BFSNodeVisitor : public daal::algorithms::tree_utils::regression::TreeNodeVisitor
{
public:
    Tree* roots;
    size_t treeId;
    std::queue<Node*> parentNodes;
    virtual bool onLeafNode(const daal::algorithms::tree_utils::regression::LeafNodeDescriptor &desc)
    {
        if(desc.level == 0)
        {
            Node* root = roots[treeId].root;
            (*(roots + treeId)).nNodes = 1;
            root->left = NULL;
            root->right = NULL;
            root->response = desc.response;
            root->featureIndex = 0;
            root->featureValue = 0;
            treeId++;
        }
        else
        {
            roots[treeId-1].nNodes++;
            Node* node = new Node(desc.response, 0, 0);

            Node* parent = parentNodes.front();
            if(parent->left == NULL)
            {
                parent->left = node;
            }
            else
            {
                parent->right = node;
                parentNodes.pop();
            }
        }
        return true;
    }

    virtual bool onSplitNode(const daal::algorithms::tree_utils::regression::SplitNodeDescriptor &desc)
    {
        if(desc.level == 0)
        {
            Node* root = roots[treeId].root;
            (*(roots + treeId)).nNodes = 1;
            root->left = NULL;
            root->right = NULL;
            root->response = 0;
            root->featureIndex = desc.featureIndex;
            root->featureValue = desc.featureValue;
            parentNodes.push(root);
            treeId++;
        }
        else
        {
            roots[treeId-1].nNodes++;
            Node* node = new Node(0,desc.featureIndex,desc.featureValue);

            Node* parent = parentNodes.front();
            if(parent->left == NULL)
            {
                parent->left = node;
            }
            else
            {
                parent->right = node;
                parentNodes.pop();
            }
            parentNodes.push(node);
        }
        return true;
    }

    BFSNodeVisitor(size_t nTrees): parentNodes()
    {
        roots = new Tree[nTrees];
        for(size_t i = 0; i < nTrees; i++)
        {
            roots[i].root = new Node;
        }
        treeId = 0;
    }
    ~BFSNodeVisitor()
    {
        if(roots)
            delete [] roots;
    }
};

training::ResultPtr trainModel();
size_t testModel(ModelPtr modelPtr);
void loadData(const std::string& fileName, NumericTablePtr& pData, NumericTablePtr& pDependentVar);
ModelPtr buildModel(Tree* trees);
Tree* traverseModel(ModelPtr m, BFSNodeVisitor& visitor);
bool buildTree(size_t treeId, Node* node, bool& isRoot, ModelBuilder& builder, const ParentPlace& parentPlace);

int main(int argc, char *argv[])
{
    checkArguments(argc, argv, 1, &trainDatasetFileName);

    /* train DAAL DF Classification model */
    training::ResultPtr trainingResult = trainModel();
    std::cout << "Predict on trained model" << std::endl;
    ModelPtr trainedModel = trainingResult->get(daal::algorithms::classifier::training::model);
    if(trainedModel.get())
        nTrees = trainedModel->numberOfTrees();
    size_t trainedAccurcy = testModel(trainedModel);

    /* traverse the trained model to get Tree representation */
    BFSNodeVisitor visitor(nTrees);
    Tree* trees = traverseModel(trainedModel, visitor);
    /* build the model by ModelBuilder from Tree */
    daal::algorithms::gbt::classification::ModelPtr builtModel = buildModel(trees);
    std::cout << "Predict on built model from input user Tree " << std::endl;
    size_t buildModelAccurcy = testModel(builtModel);
    const char* result = (trainedAccurcy == buildModelAccurcy) ? "successfully" : "not correctly";
    std::cout << "Model was built " << result << std::endl;

    return (trainedAccurcy == buildModelAccurcy) ? 0 : 1;
}

daal::algorithms::gbt::classification::ModelPtr buildModel(Tree* trees)
{
    /* create a model builder */
    const size_t nTreesInClass = nTrees/nClasses;
    ModelBuilder builder(nFeatures, nTreesInClass, nClasses);

    for(size_t i = 0; i < nTrees; i++)
    {
        const size_t nNodes = trees[i].nNodes;
        /* allocate the memory for certain tree */
        ModelBuilder::TreeId treeId = builder.createTree(nNodes, i/nTreesInClass);
        bool isRoot = true;
        /* recursive DFS traversing of certain tree with building model */
        buildTree(treeId, trees[i].root, isRoot, builder, ParentPlace(0,0));
    }

    return builder.getModel();
}

bool buildTree(size_t treeId, Node* node, bool& isRoot, ModelBuilder& builder, const ParentPlace& parentPlace)
{
    if(node->left != NULL && node->right != NULL)
    {
        if(isRoot)
        {
            ModelBuilder::NodeId parent = builder.addSplitNode(treeId, ModelBuilder::noParent, 0, node->featureIndex, node->featureValue);

            isRoot = false;
            buildTree(treeId, node->left, isRoot, builder, ParentPlace(parent,0));
            buildTree(treeId, node->right, isRoot, builder, ParentPlace(parent,1));
        }
        else
        {
            ModelBuilder::NodeId parent = builder.addSplitNode(treeId, parentPlace.parentId, parentPlace.place, node->featureIndex, node->featureValue);

            buildTree(treeId, node->left, isRoot, builder, ParentPlace(parent,0));
            buildTree(treeId, node->right, isRoot, builder, ParentPlace(parent,1));
        }
    }
    else
    {
        if(isRoot)
        {
            builder.addLeafNode(treeId, ModelBuilder::noParent, 0, node->response);
            isRoot = false;
        }
        else
        {
            builder.addLeafNode(treeId, parentPlace.parentId, parentPlace.place, node->response);
        }
    }

    return true;
}

size_t testModel(daal::algorithms::gbt::classification::ModelPtr modelPtr)
{
    /* Create Numeric Tables for testing data and ground truth values */
    NumericTablePtr testData;
    NumericTablePtr testGroundTruth;

    loadData(testDatasetFileName, testData, testGroundTruth);

    /* Create an algorithm object to predict values of decision forest classification */
    prediction::Batch<> algorithm(nClasses);

    /* Pass a testing data set and the trained model to the algorithm */
    algorithm.input.set(daal::algorithms::classifier::prediction::data, testData);
    algorithm.input.set(daal::algorithms::classifier::prediction::model, modelPtr);

    /* Predict values of decision forest classification */
    algorithm.compute();

    /* Retrieve the algorithm results */
    NumericTablePtr prediction = algorithm.getResult()->get(daal::algorithms::classifier::prediction::prediction);
    printNumericTable(prediction, "Gradient boosted trees prediction results (first 10 rows):", 10);
    printNumericTable(testGroundTruth, "Ground truth (first 10 rows):", 10);
    size_t nRows = 0;
    if(prediction.get())
        nRows = prediction->getNumberOfRows();

    size_t error = 0;
    for(size_t i = 0; i < nRows; i++)
    {
        error += prediction->getValue<float>(0,i) != testGroundTruth->getValue<float>(0,i);
    }

    std::cout << "Error: " << error << std::endl;
    return error;
}

training::ResultPtr trainModel()
{
    /* Create Numeric Tables for training data and dependent variables */
    NumericTablePtr trainData;
    NumericTablePtr trainDependentVariable;

    loadData(trainDatasetFileName, trainData, trainDependentVariable);

    /* Create an algorithm object to train the decision forest classification model */
    training::Batch<> algorithm(nClasses);

    /* Pass a training data set and dependent values to the algorithm */
    algorithm.input.set(daal::algorithms::classifier::training::data, trainData);
    algorithm.input.set(daal::algorithms::classifier::training::labels, trainDependentVariable);

    algorithm.parameter().maxIterations = maxIterations;
    algorithm.parameter().featuresPerNode = nFeatures;
    algorithm.parameter().minObservationsInLeafNode = minObservationsInLeafNode;

    /* Build the decision forest classification model */
    algorithm.compute();

    /* Retrieve the algorithm results */
    return algorithm.getResult();
}

void loadData(const std::string& fileName, NumericTablePtr& pData, NumericTablePtr& pDependentVar)
{
    /* Initialize FileDataSource<CSVFeatureManager> to retrieve the input data from a .csv file */
    FileDataSource<CSVFeatureManager> trainDataSource(fileName,
        DataSource::notAllocateNumericTable,
        DataSource::doDictionaryFromContext);

    /* Create Numeric Tables for training data and dependent variables */
    pData.reset(new HomogenNumericTable<double>(nFeatures, 0, NumericTable::notAllocate));
    pDependentVar.reset(new HomogenNumericTable<double>(1, 0, NumericTable::notAllocate));
    NumericTablePtr mergedData(new MergedNumericTable(pData, pDependentVar));

    /* Retrieve the data from input file */
    trainDataSource.loadDataBlock(mergedData.get());

    NumericTableDictionaryPtr pDictionary = pData->getDictionarySharedPtr();
    for(size_t i = 0, n = sizeof(categoricalFeaturesIndices) / sizeof(categoricalFeaturesIndices[0]); i < n; ++i)
        (*pDictionary)[categoricalFeaturesIndices[i]].featureType = data_feature_utils::DAAL_CATEGORICAL;
}

Tree* traverseModel(const daal::algorithms::gbt::classification::ModelPtr m, BFSNodeVisitor& visitor)
{
    const size_t nTrees = m->numberOfTrees();

    for(size_t i = 0; i < nTrees; ++i)
    {
        m->traverseBFS(i, visitor);
    }
    return visitor.roots;
}
Optimization Notice: 
For more complete information about compiler optimizations, see our Optimization Notice.
Select sticky button color: 
Orange (only for download buttons)