Examples with DecisionTreeNode org.knime.base.node.mine.decisiontree2.model.DecisionTreeNode used on opensource projects

Example 16 with DecisionTreeNode

use of org.knime.base.node.mine.decisiontree2.model.DecisionTreeNode in project knime-core by knime.

the class Pruner method mdlPruningRecurse.

// /**
// * The general idea is to recursively prune the children and then compare
// * the potential leaf estimated erro with the actual estimated error
// * including the length of the children.
// *
// * @param node the node to prune
// * @param zValue the z value according to which the error is estimated
// *            calculated from the confidence value
// *
// * @return the resulting description length after pruning; this value is
// *         used in higher levels of the recursion, i.e. for the parent node
// */
// private static PruningResult estimatedErrorPruningRecurse(
// final DecisionTreeNode node, final double zValue) {
// 
// // if this is a child, just return the estimated error
// if (node.isLeaf()) {
// double error = node.getEntireClassCount() - node.getOwnClassCount();
// double estimatedError =
// estimatedError(node.getEntireClassCount(), error, zValue);
// 
// return new PruningResult(estimatedError, node);
// }
// 
// // holds the estimated errors of the children
// double[] childDescriptionLength = new double[node.getChildCount()];
// DecisionTreeNodeSplit splitNode = (DecisionTreeNodeSplit)node;
// // prune all children
// DecisionTreeNode[] children = splitNode.getChildren();
// int count = 0;
// for (DecisionTreeNode childNode : children) {
// 
// PruningResult result =
// estimatedErrorPruningRecurse(childNode, zValue);
// childDescriptionLength[count] = result.getQualityValue();
// 
// // replace the child with the one from the result (could of course
// // be the same)
// splitNode.replaceChild(childNode, result.getNode());
// 
// count++;
// }
// 
// // calculate the estimated error if this would be a leaf
// double error = node.getEntireClassCount() - node.getOwnClassCount();
// double leafEstimatedError =
// estimatedError(node.getEntireClassCount(), error, zValue);
// 
// // calculate the current estimated error (sum of estimated errors of the
// // children)
// double currentEstimatedError = 0;
// for (double childDescLength : childDescriptionLength) {
// currentEstimatedError += childDescLength;
// }
// 
// // define the return node
// DecisionTreeNode returnNode = node;
// double returnEstimatedError = currentEstimatedError;
// 
// // if the possible leaf costs are smaller, replace this node
// // with a leaf (tollerance is 0.1)
// if (leafEstimatedError <= currentEstimatedError + 0.1) {
// DecisionTreeNodeLeaf newLeaf =
// new DecisionTreeNodeLeaf(node.getOwnIndex(), node
// .getMajorityClass(), node.getClassCounts());
// newLeaf.setParent((DecisionTreeNode)node.getParent());
// newLeaf.setPrefix(node.getPrefix());
// returnNode = newLeaf;
// returnEstimatedError = leafEstimatedError;
// }
// 
// return new PruningResult(returnEstimatedError, returnNode);
// }
// 
// /**
// * Prunes a {@link DecisionTree} according to the estimated error pruning
// * (Quinlan 87).
// *
// * @param decTree the decision tree to prune
// * @param confidence the confidence value according to which the error is
// *            estimated
// */
// public static void estimatedErrorPruning(final DecisionTree decTree,
// final double confidence) {
// 
// // traverse the tree depth first (in-fix)
// DecisionTreeNode root = decTree.getRootNode();
// // double zValue = xnormi(1 - confidence);
// estimatedErrorPruningRecurse(root, zValue);
// }
/**
 * The general idea is to recursively prune the children and then compare
 * the potential leaf description length with the actual length including
 * the length of the children.
 *
 * @param node the node to prune
 *
 * @return the resulting description length after pruning; this value is
 *         used in higher levels of the recursion, i.e. for the parent node
 */
private static PruningResult mdlPruningRecurse(final DecisionTreeNode node) {
    // leaf
    if (node.isLeaf()) {
        double error = node.getEntireClassCount() - node.getOwnClassCount();
        // node => 1Bit)
        return new PruningResult(error + 1.0, node);
    }
    // holds the description length of the children
    double[] childDescriptionLength = new double[node.getChildCount()];
    DecisionTreeNodeSplit splitNode = (DecisionTreeNodeSplit) node;
    // prune all children
    DecisionTreeNode[] children = splitNode.getChildren();
    int count = 0;
    for (DecisionTreeNode childNode : children) {
        PruningResult result = mdlPruningRecurse(childNode);
        childDescriptionLength[count] = result.getQualityValue();
        // replace the child with the one from the result (could of course
        // be the same)
        splitNode.replaceChild(childNode, result.getNode());
        count++;
    }
    // calculate the cost if this would be a leaf
    double leafCost = node.getEntireClassCount() - node.getOwnClassCount() + 1.0;
    // calculate the current cost including the children
    double currentCost = 1.0 + Math.log(node.getChildCount()) / Math.log(2);
    for (double childDescLength : childDescriptionLength) {
        currentCost += childDescLength;
    }
    // define the return node
    DecisionTreeNode returnNode = node;
    double returnCost = currentCost;
    // with a leaf
    if (leafCost <= currentCost) {
        DecisionTreeNodeLeaf newLeaf = new DecisionTreeNodeLeaf(node.getOwnIndex(), node.getMajorityClass(), node.getClassCounts());
        newLeaf.setParent(node.getParent());
        newLeaf.setPrefix(node.getPrefix());
        returnNode = newLeaf;
        returnCost = leafCost;
    }
    return new PruningResult(returnCost, returnNode);
}

Example 17 with DecisionTreeNode

use of org.knime.base.node.mine.decisiontree2.model.DecisionTreeNode in project knime-core by knime.

the class Pruner method mdlPruning.

/**
 * Prunes a {@link DecisionTree} according to the minimum description lenght
 * (MDL) principle.
 *
 * @param decTree the decision tree to prune
 */
public static void mdlPruning(final DecisionTree decTree) {
    // traverse the tree depth first (in-fix)
    DecisionTreeNode root = decTree.getRootNode();
    mdlPruningRecurse(root);
}

Example 18 with DecisionTreeNode

use of org.knime.base.node.mine.decisiontree2.model.DecisionTreeNode in project knime-core by knime.

the class DecTreePredictorGraphView method recreateHiLite.

private void recreateHiLite() {
    Set<RowKey> hilited = m_hiLiteHdl.getHiLitKeys();
    Set<DecisionTreeNode> toHilite = new HashSet<DecisionTreeNode>();
    DecisionTreeNode root = m_graph.getRootNode();
    List<DecisionTreeNode> toProcess = new LinkedList<DecisionTreeNode>();
    if (null != root) {
        toProcess.add(0, root);
    }
    // Traverse the tree breadth first
    while (!toProcess.isEmpty()) {
        DecisionTreeNode curr = toProcess.remove(0);
        if (hilited.containsAll(curr.coveredPattern())) {
            // hilite subtree starting from curr
            toHilite.addAll(getSubtree(curr));
        } else {
            for (int i = 0; i < curr.getChildCount(); i++) {
                toProcess.add(0, curr.getChildAt(i));
            }
        }
    }
    m_graph.hiLite(toHilite);
}

Example 19 with DecisionTreeNode

use of org.knime.base.node.mine.decisiontree2.model.DecisionTreeNode in project knime-core by knime.

the class DecTreePredictorGraphView method createTreeMenu.

/**
 * Create menu to control tree.
 *
 * @return A new JMenu with tree operation buttons
 */
private JMenu createTreeMenu() {
    final JMenu result = new JMenu("Tree");
    result.setMnemonic('T');
    Action expand = new ExpandBranchAction<DecisionTreeNode>(m_graph);
    expand.putValue(Action.NAME, "Expand Selected Branch");
    Action collapse = new CollapseBranchAction<DecisionTreeNode>(m_graph);
    collapse.putValue(Action.NAME, "Collapse Selected Branch");
    result.add(expand);
    result.add(collapse);
    return result;
}

Example 20 with DecisionTreeNode

use of org.knime.base.node.mine.decisiontree2.model.DecisionTreeNode in project knime-core by knime.

the class DecTreeLearnerGraphView2 method updateHiLite.

private void updateHiLite(final boolean state) {
    DecisionTreeNode selected = m_graph.getSelected();
    Set<RowKey> covPat = new HashSet<RowKey>();
    covPat.addAll(selected.coveredPattern());
    if (state) {
        m_hiLiteHdl.fireHiLiteEvent(covPat);
    } else {
        m_hiLiteHdl.fireUnHiLiteEvent(covPat);
    }
}