public static void main(String[] args) {
if (!parseArgs(args)) throw new IllegalArgumentException("Parse arguments failed.");
SparkConf conf = new SparkConf().setAppName("Logistic Regression with SGD");
SparkContext sc = new SparkContext(conf);
JavaRDD<String> data = sc.textFile(inputFile, 1).toJavaRDD();
JavaRDD<LabeledPoint> training =
data.map(
new Function<String, LabeledPoint>() {
public LabeledPoint call(String line) {
String[] splits = line.split(",");
double[] features = new double[3];
try {
features[0] = Double.valueOf(splits[1]);
features[1] = Double.valueOf(splits[2]);
features[2] = Double.valueOf(splits[3]);
return new LabeledPoint(Double.valueOf(splits[3]), Vectors.dense(features));
} catch (NumberFormatException e) {
return null; // Nothing to do..
}
}
})
.filter(
new Function<LabeledPoint, Boolean>() {
public Boolean call(LabeledPoint p) {
return p != null;
}
})
.cache();
LogisticRegressionModel model = lrs.run(training.rdd());
model.save(sc, outputFile);
sc.stop();
}
public void execute(Tuple input) {
LOG.info("Entered prediction bolt execute...");
String eventType = input.getStringByField("eventType");
double prediction;
if (eventType.equals("Normal")) {
double[] predictionParams = enrichEvent(input);
prediction = model.predict(Vectors.dense(predictionParams));
LOG.info("Prediction is: " + prediction);
String driverName = input.getStringByField("driverName");
String routeName = input.getStringByField("routeName");
int truckId = input.getIntegerByField("truckId");
Timestamp eventTime = (Timestamp) input.getValueByField("eventTime");
double longitude = input.getDoubleByField("longitude");
double latitude = input.getDoubleByField("latitude");
double driverId = input.getIntegerByField("driverId");
SimpleDateFormat sdf = new SimpleDateFormat();
collector.emit(
input,
new Values(
prediction == 0.0 ? "normal" : "violation",
driverName,
routeName,
driverId,
truckId,
sdf.format(new Date(eventTime.getTime())),
longitude,
latitude,
predictionParams[0] == 1 ? "Y" : "N", // driver certification status
predictionParams[1] == 1 ? "miles" : "hourly", // driver wage plan
predictionParams[2] * 100, // hours feature was scaled down by 100
predictionParams[3] * 1000, // miles feature was scaled down by 1000
predictionParams[4] == 1 ? "Y" : "N", // foggy weather
predictionParams[5] == 1 ? "Y" : "N", // rainy weather
predictionParams[6] == 1 ? "Y" : "N" // windy weather
));
if (prediction == 1.0) {
try {
writePredictionToHDFS(input, predictionParams, prediction);
} catch (Exception e) {
e.printStackTrace();
throw new RuntimeException("Couldn't write prediction to hdfs" + e);
}
}
}
// acknowledge even if there is an error
collector.ack(input);
}
/**
* This method builds a logistic regression model
*
* @param sparkContext JavaSparkContext initialized with the application
* @param modelID Model ID
* @param trainingData Training data as a JavaRDD of LabeledPoints
* @param testingData Testing data as a JavaRDD of LabeledPoints
* @param workflow Machine learning workflow
* @param mlModel Deployable machine learning model
* @param isSGD Whether the algorithm is Logistic regression with SGD
* @throws MLModelBuilderException
*/
private ModelSummary buildLogisticRegressionModel(
JavaSparkContext sparkContext,
long modelID,
JavaRDD<LabeledPoint> trainingData,
JavaRDD<LabeledPoint> testingData,
Workflow workflow,
MLModel mlModel,
SortedMap<Integer, String> includedFeatures,
boolean isSGD)
throws MLModelBuilderException {
try {
LogisticRegression logisticRegression = new LogisticRegression();
Map<String, String> hyperParameters = workflow.getHyperParameters();
LogisticRegressionModel logisticRegressionModel;
String algorithmName;
int noOfClasses = getNoOfClasses(mlModel);
if (isSGD) {
algorithmName = SUPERVISED_ALGORITHM.LOGISTIC_REGRESSION.toString();
if (noOfClasses > 2) {
throw new MLModelBuilderException(
"A binary classification algorithm cannot have more than "
+ "two distinct values in response variable.");
}
logisticRegressionModel =
logisticRegression.trainWithSGD(
trainingData,
Double.parseDouble(hyperParameters.get(MLConstants.LEARNING_RATE)),
Integer.parseInt(hyperParameters.get(MLConstants.ITERATIONS)),
hyperParameters.get(MLConstants.REGULARIZATION_TYPE),
Double.parseDouble(hyperParameters.get(MLConstants.REGULARIZATION_PARAMETER)),
Double.parseDouble(hyperParameters.get(MLConstants.SGD_DATA_FRACTION)));
} else {
algorithmName = SUPERVISED_ALGORITHM.LOGISTIC_REGRESSION_LBFGS.toString();
logisticRegressionModel =
logisticRegression.trainWithLBFGS(
trainingData, hyperParameters.get(MLConstants.REGULARIZATION_TYPE), noOfClasses);
}
// remove from cache
trainingData.unpersist();
// add test data to cache
testingData.cache();
Vector weights = logisticRegressionModel.weights();
if (!isValidWeights(weights)) {
throw new MLModelBuilderException(
"Weights of the model generated are null or infinity. [Weights] "
+ vectorToString(weights));
}
// getting scores and labels without clearing threshold to get confusion matrix
JavaRDD<Tuple2<Object, Object>> scoresAndLabelsThresholded =
logisticRegression.test(logisticRegressionModel, testingData);
MulticlassMetrics multiclassMetrics =
new MulticlassMetrics(JavaRDD.toRDD(scoresAndLabelsThresholded));
MulticlassConfusionMatrix multiclassConfusionMatrix =
getMulticlassConfusionMatrix(multiclassMetrics, mlModel);
// clearing the threshold value to get a probability as the output of the prediction
logisticRegressionModel.clearThreshold();
JavaRDD<Tuple2<Object, Object>> scoresAndLabels =
logisticRegression.test(logisticRegressionModel, testingData);
ProbabilisticClassificationModelSummary probabilisticClassificationModelSummary =
SparkModelUtils.generateProbabilisticClassificationModelSummary(
sparkContext, testingData, scoresAndLabels);
mlModel.setModel(new MLClassificationModel(logisticRegressionModel));
// remove from cache
testingData.unpersist();
List<FeatureImportance> featureWeights =
getFeatureWeights(includedFeatures, logisticRegressionModel.weights().toArray());
probabilisticClassificationModelSummary.setFeatures(
includedFeatures.values().toArray(new String[0]));
probabilisticClassificationModelSummary.setFeatureImportance(featureWeights);
probabilisticClassificationModelSummary.setAlgorithm(algorithmName);
probabilisticClassificationModelSummary.setMulticlassConfusionMatrix(
multiclassConfusionMatrix);
Double modelAccuracy = getModelAccuracy(multiclassMetrics);
probabilisticClassificationModelSummary.setModelAccuracy(modelAccuracy);
probabilisticClassificationModelSummary.setDatasetVersion(workflow.getDatasetVersion());
return probabilisticClassificationModelSummary;
} catch (Exception e) {
throw new MLModelBuilderException(
"An error occurred while building logistic regression model: " + e.getMessage(), e);
}
}
