IGNITE-11003: Added support of Random Forest

examples/src/main/java/org/apache/ignite/examples/ml/inference/spark/modelparser/RandomForestFromSparkExample.java

@@ -0,0 +1,85 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements.  See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You under the Apache License, Version 2.0
+ * (the "License"); you may not use this file except in compliance with
+ * the License.  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.ignite.examples.ml.inference.spark.modelparser;
+
+import java.io.FileNotFoundException;
+import org.apache.ignite.Ignite;
+import org.apache.ignite.IgniteCache;
+import org.apache.ignite.Ignition;
+import org.apache.ignite.examples.ml.tutorial.TitanicUtils;
+import org.apache.ignite.ml.composition.ModelsComposition;
+import org.apache.ignite.ml.math.functions.IgniteBiFunction;
+import org.apache.ignite.ml.math.primitives.vector.Vector;
+import org.apache.ignite.ml.math.primitives.vector.VectorUtils;
+import org.apache.ignite.ml.selection.scoring.evaluator.BinaryClassificationEvaluator;
+import org.apache.ignite.ml.selection.scoring.metric.Accuracy;
+import org.apache.ignite.ml.sparkmodelparser.SparkModelParser;
+import org.apache.ignite.ml.sparkmodelparser.SupportedSparkModels;
+
+/**
+ * Run Random Forest model loaded from snappy.parquet file.
+ * The snappy.parquet file was generated by Spark MLLib model.write.overwrite().save(..) operator.
+ * <p>
+ * You can change the test data used in this example and re-run it to explore this algorithm further.</p>
+ */
+public class RandomForestFromSparkExample {
+    /** Path to Spark Random Forest model. */
+    public static final String SPARK_MDL_PATH = "examples/src/main/resources/models/spark/serialized/rf/data" +
+        "/part-00000-290bdb9d-bc1b-411c-8811-c3205434f5fc-c000.snappy.parquet";
+
+    /** Run example. */
+    public static void main(String[] args) throws FileNotFoundException {
+        System.out.println();
+        System.out.println(">>> Random Forest model loaded from Spark through serialization over partitioned dataset usage example started.");
+        // Start ignite grid.
+        try (Ignite ignite = Ignition.start("examples/config/example-ignite.xml")) {
+            System.out.println(">>> Ignite grid started.");
+
+            IgniteCache<Integer, Object[]> dataCache = TitanicUtils.readPassengers(ignite);
+
+            IgniteBiFunction<Integer, Object[], Vector> featureExtractor = (k, v) -> {
+                double[] data = new double[] {(double)v[0], (double)v[5], (double)v[6]};
+                data[0] = Double.isNaN(data[0]) ? 0 : data[0];
+                data[1] = Double.isNaN(data[1]) ? 0 : data[1];
+                data[2] = Double.isNaN(data[2]) ? 0 : data[2];
+
+                return VectorUtils.of(data);
+            };
+
+            IgniteBiFunction<Integer, Object[], Double> lbExtractor = (k, v) -> (double)v[1];
+
+            ModelsComposition mdl = (ModelsComposition)SparkModelParser.parse(
+                SPARK_MDL_PATH,
+                SupportedSparkModels.RANDOM_FOREST
+            );
+
+            System.out.println(">>> Random Forest model: " + mdl.toString(true));
+
+            double accuracy = BinaryClassificationEvaluator.evaluate(
+                dataCache,
+                mdl,
+                featureExtractor,
+                lbExtractor,
+                new Accuracy<>()
+            );
+
+            System.out.println("\n>>> Accuracy " + accuracy);
+            System.out.println("\n>>> Test Error " + (1 - accuracy));
+        }
+    }
+}

examples/src/main/resources/models/spark/serialized/rf/metadata/part-00000

@@ -0,0 +1 @@
+{"class":"org.apache.spark.ml.classification.RandomForestClassificationModel","timestamp":1548169635203,"sparkVersion":"2.2.0","uid":"rfc_4627f663b8c3","paramMap":{"featureSubsetStrategy":"auto","maxMemoryInMB":256,"impurity":"gini","numTrees":200,"probabilityCol":"probability","maxDepth":10,"labelCol":"survived","maxBins":32,"subsamplingRate":1.0,"rawPredictionCol":"rawPrediction","checkpointInterval":10,"featuresCol":"features","minInstancesPerNode":1,"predictionCol":"prediction","seed":207336481,"cacheNodeIds":false,"minInfoGain":0.0},"numFeatures":3,"numClasses":2,"numTrees":200}

modules/ml/spark-model-parser/src/main/java/org/apache/ignite/ml/sparkmodelparser/SparkModelParser.java

@@ -19,12 +19,17 @@
 
 import java.io.File;
 import java.io.IOException;
+import java.util.ArrayList;
+import java.util.List;
 import java.util.Map;
 import java.util.NavigableMap;
 import java.util.TreeMap;
 import org.apache.hadoop.conf.Configuration;
 import org.apache.hadoop.fs.Path;
 import org.apache.ignite.internal.util.IgniteUtils;
+import org.apache.ignite.ml.IgniteModel;
+import org.apache.ignite.ml.composition.ModelsComposition;
+import org.apache.ignite.ml.composition.predictionsaggregator.OnMajorityPredictionsAggregator;
 import org.apache.ignite.ml.inference.Model;
 import org.apache.ignite.ml.math.primitives.vector.Vector;
 import org.apache.ignite.ml.math.primitives.vector.impl.DenseVector;
@@ -72,11 +77,52 @@ public static Model parse(String pathToMdl, SupportedSparkModels parsedSparkMdl)
                 return loadLinearSVMModel(ignitePathToMdl);
             case DECISION_TREE:
                 return loadDecisionTreeModel(ignitePathToMdl);
+            case RANDOM_FOREST:
+                return loadRandomForestModel(ignitePathToMdl);
             default:
                 throw new UnsupportedSparkModelException(ignitePathToMdl);
         }
     }
 
+    private static Model loadRandomForestModel(String pathToMdl) {
+        try (ParquetFileReader r = ParquetFileReader.open(HadoopInputFile.fromPath(new Path(pathToMdl), new Configuration()))) {
+            PageReadStore pages;
+            final MessageType schema = r.getFooter().getFileMetaData().getSchema();
+            final MessageColumnIO colIO = new ColumnIOFactory().getColumnIO(schema);
+            final Map<Integer, TreeMap<Integer, NodeData>> nodesByTreeId = new TreeMap<>();
+            while (null != (pages = r.readNextRowGroup())) {
+                final long rows = pages.getRowCount();
+                final RecordReader recordReader = colIO.getRecordReader(pages, new GroupRecordConverter(schema));
+                for (int i = 0; i < rows; i++) {
+                    final SimpleGroup g = (SimpleGroup)recordReader.read();
+                    final int treeID = g.getInteger(0, 0);
+                    final SimpleGroup nodeDataGroup = (SimpleGroup)g.getGroup(1, 0);
+                    NodeData nodeData = extractNodeDataFromParquetRow(nodeDataGroup);
+
+                    if (nodesByTreeId.containsKey(treeID)) {
+                        Map<Integer, NodeData> nodesByNodeId = nodesByTreeId.get(treeID);
+                        nodesByNodeId.put(nodeData.id, nodeData);
+                    }
+                    else {
+                        TreeMap<Integer, NodeData> nodesByNodeId = new TreeMap<>();
+                        nodesByNodeId.put(nodeData.id, nodeData);
+                        nodesByTreeId.put(treeID, nodesByNodeId);
+                    }
+                }
+            }
+
+            final List<IgniteModel<Vector, Double>> models = new ArrayList<>();
+            nodesByTreeId.forEach((key, nodes) -> models.add(buildDecisionTreeModel(nodes)));
+
+            return new ModelsComposition(models, new OnMajorityPredictionsAggregator());
+        }
+        catch (IOException e) {
+            System.out.println("Error reading parquet file.");
+            e.printStackTrace();
+        }
+        return null;
+    }
+
     /**
      * Load Decision Tree model.
      *
@@ -111,7 +157,7 @@ private static Model loadDecisionTreeModel(String pathToMdl) {
      *
      * @param nodes The sorted map of nodes.
      */
-    private static Model buildDecisionTreeModel(Map<Integer, NodeData> nodes) {
+    private static DecisionTreeNode buildDecisionTreeModel(Map<Integer, NodeData> nodes) {
         DecisionTreeNode mdl = null;
         if (!nodes.isEmpty()) {
             NodeData rootNodeData = (NodeData)((NavigableMap)nodes).firstEntry().getValue();

modules/ml/spark-model-parser/src/main/java/org/apache/ignite/ml/sparkmodelparser/SupportedSparkModels.java

@@ -33,5 +33,8 @@ public enum SupportedSparkModels {
     DECISION_TREE,
 
     /** Support Vector Machine . */
-    LINEAR_SVM
+    LINEAR_SVM,
+
+    /** Random forest. */
+    RANDOM_FOREST
 }