private static String train(
SQLContext sqlContext, DataFrame positives, DataFrame negatives, String modelFileName) {
// combine data sets
DataFrame all = positives.unionAll(negatives);
// split into training and test sets
DataFrame[] split = all.randomSplit(new double[] {.80, .20}, 1);
DataFrame training = split[0].cache();
DataFrame test = split[1].cache();
// fit logistic regression model
PipelineModel model = fitLogisticRegressionModel(training);
try {
ObjectSerializer.serialize(model, modelFileName);
} catch (IOException e1) {
// TODO Auto-generated catch block
e1.printStackTrace();
}
// predict on training data to evaluate goodness of fit
DataFrame trainingResults = model.transform(training).cache();
// predict on test set to evaluate goodness of fit
DataFrame testResults = model.transform(test).cache();
// predict on unassigned data mentions
StringBuilder sb = new StringBuilder();
sb.append(getMetrics(trainingResults, "Training\n"));
sb.append(getMetrics(testResults, "Testing\n"));
return sb.toString();
}
public static void main(String[] args) {
SparkConf conf = new SparkConf().setAppName("JavaDecisionTreeRegressionExample");
JavaSparkContext jsc = new JavaSparkContext(conf);
SQLContext sqlContext = new SQLContext(jsc);
// $example on$
// Load the data stored in LIBSVM format as a DataFrame.
DataFrame data = sqlContext.read().format("libsvm").load("data/mllib/sample_libsvm_data.txt");
// Automatically identify categorical features, and index them.
// Set maxCategories so features with > 4 distinct values are treated as continuous.
VectorIndexerModel featureIndexer =
new VectorIndexer()
.setInputCol("features")
.setOutputCol("indexedFeatures")
.setMaxCategories(4)
.fit(data);
// Split the data into training and test sets (30% held out for testing)
DataFrame[] splits = data.randomSplit(new double[] {0.7, 0.3});
DataFrame trainingData = splits[0];
DataFrame testData = splits[1];
// Train a DecisionTree model.
DecisionTreeRegressor dt = new DecisionTreeRegressor().setFeaturesCol("indexedFeatures");
// Chain indexer and tree in a Pipeline
Pipeline pipeline = new Pipeline().setStages(new PipelineStage[] {featureIndexer, dt});
// Train model. This also runs the indexer.
PipelineModel model = pipeline.fit(trainingData);
// Make predictions.
DataFrame predictions = model.transform(testData);
// Select example rows to display.
predictions.select("label", "features").show(5);
// Select (prediction, true label) and compute test error
RegressionEvaluator evaluator =
new RegressionEvaluator()
.setLabelCol("label")
.setPredictionCol("prediction")
.setMetricName("rmse");
double rmse = evaluator.evaluate(predictions);
System.out.println("Root Mean Squared Error (RMSE) on test data = " + rmse);
DecisionTreeRegressionModel treeModel = (DecisionTreeRegressionModel) (model.stages()[1]);
System.out.println("Learned regression tree model:\n" + treeModel.toDebugString());
// $example off$
}
@Override
public int run(SparkConf conf, CommandLine cli) throws Exception {
long startMs = System.currentTimeMillis();
conf.set("spark.ui.enabled", "false");
JavaSparkContext jsc = new JavaSparkContext(conf);
SQLContext sqlContext = new SQLContext(jsc);
long diffMs = (System.currentTimeMillis() - startMs);
System.out.println(">> took " + diffMs + " ms to create SQLContext");
Map<String, String> options = new HashMap<>();
options.put("zkhost", "localhost:9983");
options.put("collection", "ml20news");
options.put("query", "content_txt:[* TO *]");
options.put("fields", "content_txt");
DataFrame solrData = sqlContext.read().format("solr").options(options).load();
DataFrame sample = solrData.sample(false, 0.1d, 5150).select("content_txt");
List<Row> rows = sample.collectAsList();
System.out.println(">> loaded " + rows.size() + " docs to classify");
StructType schema = sample.schema();
CrossValidatorModel cvModel = CrossValidatorModel.load("ml-pipeline-model");
PipelineModel bestModel = (PipelineModel) cvModel.bestModel();
int r = 0;
startMs = System.currentTimeMillis();
for (Row next : rows) {
Row oneRow = RowFactory.create(next.getString(0));
DataFrame oneRowDF =
sqlContext.createDataFrame(Collections.<Row>singletonList(oneRow), schema);
DataFrame scored = bestModel.transform(oneRowDF);
Row scoredRow = scored.collect()[0];
String predictedLabel = scoredRow.getString(scoredRow.fieldIndex("predictedLabel"));
// an acutal app would save the predictedLabel
// System.out.println(">> for row["+r+"], model returned "+scoredRows.length+" rows,
// "+scoredRows[0]);
r++;
}
diffMs = (System.currentTimeMillis() - startMs);
System.out.println(">> took " + diffMs + " ms to score " + rows.size() + " docs");
return 0;
}
