/java-local-outlier-factor

Package implements a number local outlier factor algorithms for outlier detection and finding anomalous data

Primary LanguageJavaMIT LicenseMIT

java-local-outlier-factor

Package implements a number local outlier factor algorithms for outlier detection and finding anomalous data

Build Status Coverage Status

Features

  • LOF
  • LDOF (Local Density Outlier Factor)
  • LOCI (Local outlier correlation integral)
  • CBLOF (Cluster-based LOF)

Install

Add the following dependency to your POM file:

<dependency>
  <groupId>com.github.chen0040</groupId>
  <artifactId>java-local-outlier-factor</artifactId>
  <version>1.0.4</version>
</dependency>

Usage

The anomaly detection algorithms takes data that is prepared and stored in a data frame (Please refers to this link on how to create a data frame from file or from scratch)

All LOF algorithms variants use unsupervised-learning for training.

The following method trains an algorithm:

lof.fitAndTransform(dataFrame);

The following method returns true if the dataRow (which is a row in a data frame) taken in is an outlier:

boolean isOutlier = lof.isAnomaly(dataRow);

Local Outlier Factor (LOF)

To create and train the LOF, run the following code:

LOF method = new LOF();
method.setMinPtsLB(3);
method.setMinPtsUB(15);
method.setThreshold(0.2);
DataFrame resultantTrainedData = method.fitAndTransform(trainingData);
System.out.println(resultantTrainedData.head(10));

To test the trained method on new data, run:

boolean outlier = method.isAnomaly(dataRow);

Cluster-Based Local Outlier Factor (CBLOF)

The create and train the LOF, run the following code:

CBLOF method = new CBLOF();
DataFrame resultantTrainedData = method.fitAndTransform(trainingData);
System.out.println(resultantTrainedData.head(10));

To test the trained method on new data, run:

boolean outlier = method.isAnomaly(dataRow);

The problem that we will be using as demo as the following anomaly detection problem:

scki-learn example for one-class

LOF

Below is the sample code which illustrates how to use LOF to detect outliers in the above problem:

DataQuery.DataFrameQueryBuilder schema = DataQuery.blank()
      .newInput("c1")
      .newInput("c2")
      .newOutput("anomaly")
      .end();

Sampler.DataSampleBuilder negativeSampler = new Sampler()
      .forColumn("c1").generate((name, index) -> randn() * 0.3 + (index % 2 == 0 ? -2 : 2))
      .forColumn("c2").generate((name, index) -> randn() * 0.3 + (index % 2 == 0 ? -2 : 2))
      .forColumn("anomaly").generate((name, index) -> 0.0)
      .end();

Sampler.DataSampleBuilder positiveSampler = new Sampler()
      .forColumn("c1").generate((name, index) -> rand(-4, 4))
      .forColumn("c2").generate((name, index) -> rand(-4, 4))
      .forColumn("anomaly").generate((name, index) -> 1.0)
      .end();

DataFrame data = schema.build();

data = negativeSampler.sample(data, 20);
data = positiveSampler.sample(data, 20);

System.out.println(data.head(10));

LOF method = new LOF();
method.setParallel(true);
method.setMinPtsLB(3);
method.setMinPtsUB(10);
method.setThreshold(0.5);
DataFrame learnedData = method.fitAndTransform(data);

BinaryClassifierEvaluator evaluator = new BinaryClassifierEvaluator();

for(int i = 0; i < learnedData.rowCount(); ++i){
 boolean predicted = learnedData.row(i).categoricalTarget().equals("1");
 boolean actual = data.row(i).target() == 1.0;
 evaluator.evaluate(actual, predicted);
 logger.info("predicted: {}\texpected: {}", predicted, actual);
}

Cluster-Based LOF

Below is the sample code which illustrates how to use CBLOF to detect outliers in the above problem:

DataQuery.DataFrameQueryBuilder schema = DataQuery.blank()
      .newInput("c1")
      .newInput("c2")
      .newOutput("anomaly")
      .end();

Sampler.DataSampleBuilder negativeSampler = new Sampler()
      .forColumn("c1").generate((name, index) -> randn() * 0.3 + (index % 2 == 0 ? -2 : 2))
      .forColumn("c2").generate((name, index) -> randn() * 0.3 + (index % 2 == 0 ? -2 : 2))
      .forColumn("anomaly").generate((name, index) -> 0.0)
      .end();

Sampler.DataSampleBuilder positiveSampler = new Sampler()
      .forColumn("c1").generate((name, index) -> rand(-4, 4))
      .forColumn("c2").generate((name, index) -> rand(-4, 4))
      .forColumn("anomaly").generate((name, index) -> 1.0)
      .end();

DataFrame data = schema.build();

data = negativeSampler.sample(data, 200);
data = positiveSampler.sample(data, 200);

System.out.println(data.head(10));


CBLOF method = new CBLOF();
method.setParallel(false);
DataFrame learnedData = method.fitAndTransform(data);

BinaryClassifierEvaluator evaluator = new BinaryClassifierEvaluator();

for(int i = 0; i < learnedData.rowCount(); ++i){
 boolean predicted = learnedData.row(i).categoricalTarget().equals("1");
 boolean actual = data.row(i).target() == 1.0;
 evaluator.evaluate(actual, predicted);
 logger.info("predicted: {}\texpected: {}", predicted, actual);
}

evaluator.report();

LDOF

Below is the sample code which illustrates how to use LDOF to detect outliers in the above problem:

DataQuery.DataFrameQueryBuilder schema = DataQuery.blank()
      .newInput("c1")
      .newInput("c2")
      .newOutput("anomaly")
      .end();

Sampler.DataSampleBuilder negativeSampler = new Sampler()
      .forColumn("c1").generate((name, index) -> randn() * 0.3 + (index % 2 == 0 ? -2 : 2))
      .forColumn("c2").generate((name, index) -> randn() * 0.3 + (index % 2 == 0 ? -2 : 2))
      .forColumn("anomaly").generate((name, index) -> 0.0)
      .end();

Sampler.DataSampleBuilder positiveSampler = new Sampler()
      .forColumn("c1").generate((name, index) -> rand(-4, 4))
      .forColumn("c2").generate((name, index) -> rand(-4, 4))
      .forColumn("anomaly").generate((name, index) -> 1.0)
      .end();

DataFrame data = schema.build();

data = negativeSampler.sample(data, 20);
data = positiveSampler.sample(data, 20);

System.out.println(data.head(10));

LDOF method = new LDOF();
DataFrame learnedData = method.fitAndTransform(data);

BinaryClassifierEvaluator evaluator = new BinaryClassifierEvaluator();
for(int i = 0; i < learnedData.rowCount(); ++i) {
 boolean predicted = learnedData.row(i).categoricalTarget().equals("1");
 boolean actual = data.row(i).target() == 1.0;

 evaluator.evaluate(actual, predicted);
 logger.info("predicted: {}\texpected: {}", predicted, actual);
}

evaluator.report();

LOCI

Below is the sample code which illustrates how to use LOCI to detect outliers in the above problem:

DataQuery.DataFrameQueryBuilder schema = DataQuery.blank()
      .newInput("c1")
      .newInput("c2")
      .newOutput("anomaly")
      .end();

Sampler.DataSampleBuilder negativeSampler = new Sampler()
      .forColumn("c1").generate((name, index) -> randn() * 0.3 + (index % 2 == 0 ? -2 : 2))
      .forColumn("c2").generate((name, index) -> randn() * 0.3 + (index % 2 == 0 ? -2 : 2))
      .forColumn("anomaly").generate((name, index) -> 0.0)
      .end();

Sampler.DataSampleBuilder positiveSampler = new Sampler()
      .forColumn("c1").generate((name, index) -> rand(-4, 4))
      .forColumn("c2").generate((name, index) -> rand(-4, 4))
      .forColumn("anomaly").generate((name, index) -> 1.0)
      .end();

DataFrame data = schema.build();

data = negativeSampler.sample(data, 20);
data = positiveSampler.sample(data, 20);

System.out.println(data.head(10));

LOCI method = new LOCI();
method.setAlpha(0.5);
method.setKSigma(3);
DataFrame learnedData = method.fitAndTransform(data);

BinaryClassifierEvaluator evaluator = new BinaryClassifierEvaluator();

for(int i = 0; i < learnedData.rowCount(); ++i){
 boolean predicted = learnedData.row(i).categoricalTarget().equals("1");
 boolean actual = data.row(i).target() == 1.0;
 evaluator.evaluate(actual, predicted);
 logger.info("predicted: {}\texpected: {}", predicted, actual);
}