Simple workflow API for MOA.
Makes use of the reactive API that Java introduced with Java 9:
https://community.oracle.com/docs/DOC-1006738
The following operators are currently available:
Package: com.github.fracpete.moaflow.source
InstanceSource
- generates instances using a stream generator
Package: com.github.fracpete.moaflow.transformer
EvaluatorClassifier
- evaluates a classifierEvaluatorRegressor
- evaluates a regressorInstanceFilter
- for filtering a data streamTrainClassifier
- for training a classifierTrainRegressor
- for training a regressor
Package: com.github.fracpete.moaflow.sink
Console
- simply outputs any objects it receives viatoString()
DrawTable
- draws a tableMeasurementPlot
- for plotting a single measurementMeasurementsToCSV
- stores the measurements from an evaluation as CSVMeasurementTableSawPlot
- plots realtime chart on web-browser using Jupyter Notebook and IJavaOutputLearningCurve
- outputs the whole learning curve of MOA in a table formatWriteModel
- stores a model on disk
- ClassificationCsvAndTrain -- evaluates a classifier, outputs the statistics in a CSV, also trains the model and stores it on disk
- ClassificationPlot -- evaluates a classifier and plots a statistic.
- RegressionCsvAndTrain -- filters the data, evaluates a regressor, outputs the statistics in a CSV, also trains the model and stores it on disk
- RegressionPlot -- evaluates a regressor and plots a statistic.
- RegressionConstructSaveLoadRun -- constructs, saves, loads and runs a flow that evaluates a regressor, outputs the statistics in a CSV, also trains the model and stores it on disk
- ClassificationVariousSchemaTablePlot -- uses different evaluation schemes, output a table of learning curves and plots a statistic with TableSaw (only works with IJava and Jupyter Notebook)
moa-flow-core
- core functionalitymoa-flow-examples
- example codemoa-flow-app
- meta-module for generating binary .zip and .deb package
-
compile
mvn clean install -DskipTests=true
-
build .deb package
mvn clean install deb:package -DskipTests=true