/*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
/*
* PMMLFactory.java
* Copyright (C) 2008-2012 University of Waikato, Hamilton, New Zealand
*
*/
package weka.core.pmml;
import java.io.BufferedInputStream;
import java.io.BufferedOutputStream;
import java.io.File;
import java.io.FileInputStream;
import java.io.FileOutputStream;
import java.io.InputStream;
import java.io.ObjectOutputStream;
import java.io.OutputStream;
import java.util.ArrayList;
import javax.xml.parsers.DocumentBuilder;
import javax.xml.parsers.DocumentBuilderFactory;
import org.w3c.dom.Document;
import org.w3c.dom.Element;
import org.w3c.dom.Node;
import org.w3c.dom.NodeList;
import weka.classifiers.AbstractClassifier;
import weka.classifiers.pmml.consumer.GeneralRegression;
import weka.classifiers.pmml.consumer.NeuralNetwork;
import weka.classifiers.pmml.consumer.PMMLClassifier;
import weka.classifiers.pmml.consumer.Regression;
import weka.classifiers.pmml.consumer.RuleSetModel;
import weka.classifiers.pmml.consumer.SupportVectorMachineModel;
import weka.classifiers.pmml.consumer.TreeModel;
import weka.core.Attribute;
import weka.core.Instance;
import weka.core.Instances;
import weka.core.Utils;
import weka.gui.Logger;
/**
* This class is a factory class for reading/writing PMML models
*
* @author Mark Hall (mhall{[at]}pentaho{[dot]}com)
* @version $Revision: 8034 $
*/
public class PMMLFactory {
/** for serialization */
protected enum ModelType {
UNKNOWN_MODEL ("unknown"),
REGRESSION_MODEL ("Regression"),
GENERAL_REGRESSION_MODEL ("GeneralRegression"),
NEURAL_NETWORK_MODEL ("NeuralNetwork"),
TREE_MODEL ("TreeModel"),
RULESET_MODEL("RuleSetModel"),
SVM_MODEL ("SupportVectorMachineModel");
private final String m_stringVal;
ModelType(String name) {
m_stringVal = name;
}
public String toString() {
return m_stringVal;
}
}
/**
* Read and return a PMML model.
*
* @param filename the name of the file to read from
* @return a PMML model
* @throws Exception if there is a problem while reading the file
*/
public static PMMLModel getPMMLModel(String filename) throws Exception {
return getPMMLModel(filename, null);
}
/**
* Read and return a PMML model.
*
* @param file a <code>File</code> to read from
* @return a PMML model
* @throws Exception if there is a problem while reading the file
*/
public static PMMLModel getPMMLModel(File file) throws Exception {
return getPMMLModel(file, null);
}
/**
* Read and return a PMML model.
*
* @param stream the <code>InputStream</code> to read from
* @return a PMML model
* @throws Exception if there is a problem while reading from the stream
*/
public static PMMLModel getPMMLModel(InputStream stream) throws Exception {
return getPMMLModel(stream, null);
}
/**
* Read and return a PMML model.
*
* @param filename the name of the file to read from
* @param log the logging object to use (or null if none is to be used)
* @return a PMML model
* @throws Exception if there is a problem while reading the file
*/
public static PMMLModel getPMMLModel(String filename, Logger log) throws Exception {
return getPMMLModel(new File(filename), log);
}
/**
* Read and return a PMML model.
*
* @param file a <code>File</code> to read from
* @param log the logging object to use (or null if none is to be used)
* @return a PMML model
* @throws Exception if there is a problem while reading the file
*/
public static PMMLModel getPMMLModel(File file, Logger log) throws Exception {
return getPMMLModel(new BufferedInputStream(new FileInputStream(file)), log);
}
private static boolean isPMML(Document doc) {
NodeList tempL = doc.getElementsByTagName("PMML");
if (tempL.getLength() == 0) {
return false;
}
return true;
}
/**
* Read and return a PMML model.
*
* @param stream the <code>InputStream</code> to read from
* @param log the logging object to use (or null if none is to be used)
* @return a PMML model
* @throws Exception if there is a problem while reading from the stream
*/
public static PMMLModel getPMMLModel(InputStream stream, Logger log) throws Exception {
DocumentBuilderFactory dbf = DocumentBuilderFactory.newInstance();
DocumentBuilder db = dbf.newDocumentBuilder();
Document doc = db.parse(stream);
stream.close();
doc.getDocumentElement().normalize();
if (!isPMML(doc)) {
throw new IllegalArgumentException("[PMMLFactory] Source is not a PMML file!!");
}
// System.out.println("Root element " + doc.getDocumentElement().getNodeName());
Instances dataDictionary = getDataDictionaryAsInstances(doc);
TransformationDictionary transDict = getTransformationDictionary(doc, dataDictionary);
ModelType modelType = getModelType(doc);
if (modelType == ModelType.UNKNOWN_MODEL) {
throw new Exception("Unsupported PMML model type");
}
Element model = getModelElement(doc, modelType);
// Construct mining schema and meta data
MiningSchema ms = new MiningSchema(model, dataDictionary, transDict);
//System.out.println(ms);
//System.exit(1);
// Instances miningSchema = getMiningSchemaAsInstances(model, dataDictionary);
PMMLModel theModel = getModelInstance(doc, modelType, model, dataDictionary, ms);
if (log != null) {
theModel.setLog(log);
}
return theModel;
}
/**
* Get the transformation dictionary (if there is one).
*
* @param doc the Document containing the PMML model
* @param dataDictionary the data dictionary as an Instances object
* @return the transformation dictionary or null if there is none defined in
* the Document
* @throws Exception if there is a problem getting the transformation
* dictionary
*/
protected static TransformationDictionary getTransformationDictionary(Document doc,
Instances dataDictionary) throws Exception {
TransformationDictionary transDict = null;
NodeList transL = doc.getElementsByTagName("TransformationDictionary");
// should be of size 0 or 1
if (transL.getLength() > 0) {
Node transNode = transL.item(0);
if (transNode.getNodeType() == Node.ELEMENT_NODE) {
transDict = new TransformationDictionary((Element)transNode, dataDictionary);
}
}
return transDict;
}
/**
* Serialize a <code>PMMLModel</code> object that encapsulates a PMML model
*
* @param model the <code>PMMLModel</code> to serialize
* @param filename the name of the file to save to
* @throws Exception if something goes wrong during serialization
*/
public static void serializePMMLModel(PMMLModel model, String filename)
throws Exception {
serializePMMLModel(model, new File(filename));
}
/**
* Serialize a <code>PMMLModel</code> object that encapsulates a PMML model
*
* @param model the <code>PMMLModel</code> to serialize
* @param file the <code>File</code> to save to
* @throws Exception if something goes wrong during serialization
*/
public static void serializePMMLModel(PMMLModel model, File file)
throws Exception {
serializePMMLModel(model, new BufferedOutputStream(new FileOutputStream(file)));
}
/**
* Serialize a <code>PMMLModel</code> object that encapsulates a PMML model
*
* @param model the <code>PMMLModel</code> to serialize
* @param stream the <code>OutputStream</code> to serialize to
* @throws Exception if something goes wrong during serialization
*/
public static void serializePMMLModel(PMMLModel model, OutputStream stream)
throws Exception {
ObjectOutputStream oo = new ObjectOutputStream(stream);
Instances header = model.getMiningSchema().getFieldsAsInstances();
oo.writeObject(header);
oo.writeObject(model);
oo.flush();
oo.close();
}
/**
* Get an instance of a PMMLModel from the supplied Document
*
* @param doc the Document holding the pmml
* @param modelType the type of model
* @param model the Element encapsulating the model part of the Document
* @param dataDictionary the data dictionary as an Instances object
* @param miningSchema the mining schema
* @return a PMMLModel object
* @throws Exception if there is a problem constructing the model or
* if the model type is not supported
*/
protected static PMMLModel getModelInstance(Document doc,
ModelType modelType, Element model,
Instances dataDictionary,
MiningSchema miningSchema) throws Exception {
PMMLModel pmmlM = null;
switch (modelType) {
case REGRESSION_MODEL:
pmmlM = new Regression(model, dataDictionary, miningSchema);
//System.out.println(pmmlM);
break;
case GENERAL_REGRESSION_MODEL:
pmmlM = new GeneralRegression(model, dataDictionary, miningSchema);
//System.out.println(pmmlM);
break;
case NEURAL_NETWORK_MODEL:
pmmlM = new NeuralNetwork(model, dataDictionary, miningSchema);
break;
case TREE_MODEL:
pmmlM = new TreeModel(model, dataDictionary, miningSchema);
break;
case RULESET_MODEL:
pmmlM = new RuleSetModel(model, dataDictionary, miningSchema);
break;
case SVM_MODEL:
pmmlM = new SupportVectorMachineModel(model, dataDictionary, miningSchema);
break;
default:
throw new Exception("[PMMLFactory] Unknown model type!!");
}
pmmlM.setPMMLVersion(doc);
pmmlM.setCreatorApplication(doc);
return pmmlM;
}
/**
* Get the type of model
*
* @param doc the Document encapsulating the pmml
* @return the type of model
*/
protected static ModelType getModelType(Document doc) {
NodeList temp = doc.getElementsByTagName("RegressionModel");
if (temp.getLength() > 0) {
return ModelType.REGRESSION_MODEL;
}
temp = doc.getElementsByTagName("GeneralRegressionModel");
if (temp.getLength() > 0) {
return ModelType.GENERAL_REGRESSION_MODEL;
}
temp = doc.getElementsByTagName("NeuralNetwork");
if (temp.getLength() > 0) {
return ModelType.NEURAL_NETWORK_MODEL;
}
temp = doc.getElementsByTagName("TreeModel");
if (temp.getLength() > 0) {
return ModelType.TREE_MODEL;
}
temp = doc.getElementsByTagName("RuleSetModel");
if (temp.getLength() > 0) {
return ModelType.RULESET_MODEL;
}
temp = doc.getElementsByTagName("SupportVectorMachineModel");
if (temp.getLength() > 0) {
return ModelType.SVM_MODEL;
}
return ModelType.UNKNOWN_MODEL;
}
/**
* Get the Element that contains the pmml model
*
* @param doc the Document encapsulating the pmml
* @param modelType the type of model
* @throws Exception if the model type is unsupported/unknown
*/
protected static Element getModelElement(Document doc, ModelType modelType)
throws Exception {
NodeList temp = null;
Element model = null;
switch (modelType) {
case REGRESSION_MODEL:
temp = doc.getElementsByTagName("RegressionModel");
break;
case GENERAL_REGRESSION_MODEL:
temp = doc.getElementsByTagName("GeneralRegressionModel");
break;
case NEURAL_NETWORK_MODEL:
temp = doc.getElementsByTagName("NeuralNetwork");
break;
case TREE_MODEL:
temp = doc.getElementsByTagName("TreeModel");
break;
case RULESET_MODEL:
temp = doc.getElementsByTagName("RuleSetModel");
break;
case SVM_MODEL:
temp = doc.getElementsByTagName("SupportVectorMachineModel");
break;
default:
throw new Exception("[PMMLFactory] unknown/unsupported model type.");
}
if (temp != null && temp.getLength() > 0) {
Node modelNode = temp.item(0);
if (modelNode.getNodeType() == Node.ELEMENT_NODE) {
model = (Element)modelNode;
}
}
return model;
}
/**
* Get the mining schema as an Instances object
*
* @param model the Element containing the pmml model
* @param dataDictionary the data dictionary as an Instances object
* @return the mining schema as an Instances object
* @throws Exception if something goes wrong during reading the mining schema
* @deprecated Use the MiningSchema class instead
*/
protected static Instances getMiningSchemaAsInstances(Element model,
Instances dataDictionary)
throws Exception {
ArrayList<Attribute> attInfo = new ArrayList<Attribute>();
NodeList fieldList = model.getElementsByTagName("MiningField");
int classIndex = -1;
int addedCount = 0;
for (int i = 0; i < fieldList.getLength(); i++) {
Node miningField = fieldList.item(i);
if (miningField.getNodeType() == Node.ELEMENT_NODE) {
Element miningFieldEl = (Element)miningField;
String name = miningFieldEl.getAttribute("name");
String usage = miningFieldEl.getAttribute("usageType");
// TO-DO: also missing value replacement etc.
// find this attribute in the dataDictionary
Attribute miningAtt = dataDictionary.attribute(name);
if (miningAtt != null) {
if (usage.length() == 0 || usage.equals("active") || usage.equals("predicted")) {
attInfo.add(miningAtt);
addedCount++;
}
if (usage.equals("predicted")) {
classIndex = addedCount - 1;
}
} else {
throw new Exception("Can't find mining field: " + name
+ " in the data dictionary.");
}
}
}
Instances insts = new Instances("miningSchema", attInfo, 0);
// System.out.println(insts);
if (classIndex != -1) {
insts.setClassIndex(classIndex);
}
return insts;
}
/**
* Get the data dictionary as an Instances object
*
* @param doc the Document encapsulating the pmml
* @return the data dictionary as an Instances object
* @throws Exception if there are fields that are not continuous,
* ordinal or categorical in the data dictionary
*/
protected static Instances getDataDictionaryAsInstances(Document doc)
throws Exception {
// TO-DO: definition of missing values (see below)
ArrayList<Attribute> attInfo = new ArrayList<Attribute>();
NodeList dataDictionary = doc.getElementsByTagName("DataField");
for (int i = 0; i < dataDictionary.getLength(); i++) {
Node dataField = dataDictionary.item(i);
if (dataField.getNodeType() == Node.ELEMENT_NODE) {
Element dataFieldEl = (Element)dataField;
String name = dataFieldEl.getAttribute("name");
String type = dataFieldEl.getAttribute("optype");
Attribute tempAtt = null;
if (name != null && type != null) {
if (type.equals("continuous")) {
tempAtt = new Attribute(name);
} else if (type.equals("categorical") || type.equals("ordinal")) {
NodeList valueList = dataFieldEl.getElementsByTagName("Value");
if (valueList == null || valueList.getLength() == 0) {
// assume that categorical values will be revealed in the actual model.
// Create a string attribute for now
ArrayList<String> nullV = null;
tempAtt = new Attribute(name, (ArrayList<String>)nullV);
} else {
// add the values (if defined as "valid")
ArrayList<String> valueVector = new ArrayList<String>();
for (int j = 0; j < valueList.getLength(); j++) {
Node val = valueList.item(j);
if (val.getNodeType() == Node.ELEMENT_NODE) {
// property is optional (default value is "valid")
String property = ((Element)val).getAttribute("property");
if (property == null || property.length() == 0 || property.equals("valid")) {
String value = ((Element)val).getAttribute("value");
valueVector.add(value);
} else {
// Just ignore invalid or missing value definitions for now...
// TO-DO: implement Value meta data with missing/invalid value defs.
}
}
}
tempAtt = new Attribute(name, valueVector);
}
} else {
throw new Exception("[PMMLFactory] can't handle " + type + "attributes.");
}
attInfo.add(tempAtt);
}
}
}
// TO-DO: check whether certain values are declared to represent
// missing or invalid values (applies to both categorical and continuous
// attributes
// create the Instances structure
Instances insts = new Instances("dataDictionary", attInfo, 0);
// System.out.println(insts);
return insts;
}
public static String applyClassifier(PMMLModel model, Instances test) throws Exception {
StringBuffer buff = new StringBuffer();
if (!(model instanceof PMMLClassifier)) {
throw new Exception("PMML model is not a classifier!");
}
double[] preds = null;
PMMLClassifier classifier = (PMMLClassifier)model;
for (int i = 0; i < test.numInstances(); i++) {
buff.append("Actual: ");
Instance temp = test.instance(i);
if (temp.classAttribute().isNumeric()) {
buff.append(temp.value(temp.classIndex()) + " ");
} else {
buff.append(temp.classAttribute().value((int)temp.value(temp.classIndex())) + " ");
}
preds = classifier.distributionForInstance(temp);
buff.append(" Predicted: ");
for (int j = 0; j < preds.length; j++) {
buff.append("" + preds[j] + " ");
}
buff.append("\n");
}
return buff.toString();
}
private static class PMMLClassifierRunner extends AbstractClassifier {
public double[] distributionForInstance(Instance test) throws Exception {
throw new Exception("Don't call this method!!");
}
public void buildClassifier(Instances instances) throws Exception {
throw new Exception("Don't call this method!!");
}
public String getRevision() {
return weka.core.RevisionUtils.extract("$Revision: 8034 $");
}
public void evaluatePMMLClassifier(String[] options) {
runClassifier(this, options);
}
}
public static void main(String[] args) {
try {
String[] optionsTmp = new String[args.length];
for (int i = 0; i < args.length; i++) {
optionsTmp[i] = args[i];
}
String pmmlFile = Utils.getOption('l', optionsTmp);
if (pmmlFile.length() == 0) {
throw new Exception("[PMMLFactory] must specify a PMML file using the -l option.");
}
// see if it is supported before going any further
getPMMLModel(pmmlFile, null);
PMMLClassifierRunner pcr = new PMMLClassifierRunner();
pcr.evaluatePMMLClassifier(args);
/*PMMLModel model = getPMMLModel(args[0], null);
System.out.println(model);
if (args.length == 2) {
// load an arff file
Instances testData = new Instances(new java.io.BufferedReader(new java.io.FileReader(args[1])));
Instances miningSchemaI = model.getMiningSchema().getFieldsAsInstances();
if (miningSchemaI.classIndex() >= 0) {
String className = miningSchemaI.classAttribute().name();
for (int i = 0; i < testData.numAttributes(); i++) {
if (testData.attribute(i).name().equals(className)) {
testData.setClassIndex(i);
System.out.println("Found class " + className + " in test data.");
break;
}
}
}
System.out.println(applyClassifier(model, testData));
}*/
} catch (Exception ex) {
ex.printStackTrace();
}
}
}