Skip to content
This repository has been archived by the owner on Jul 20, 2020. It is now read-only.

paypal/yurita

Repository files navigation

logo

Yurita

Join the chat at https://gitter.im/pp-yurita Build Status Codacy Badge License Documentation Status

Yurita is an open source project for developing large scale anomaly detection models Site

Getting Started

Documentation

Documentation on Yurita's architecture, statistical models available, anomaly detection pipeline/data flow, etc can be found here: https://yurita.readthedocs.io/en/latest/

Build from source

foo@bar:~/yurita$ ./gradlew clean build
foo@bar:~/yurita$ ./gradlew publishToMavenLocal

Install from Maven Central

Please build the project from source at this time or try our dockerized Yurita demo application to build automatically as we make the project jar available on Maven Central in upcoming few days.

<dependency>
    <groupId>io.github.paypal</groupId>
    <artifactId>yurita</artifactId>
    <version>1.0.0</version>
</dependency>

Other Required Dependencies:

<dependency>
    <groupId>org.apache.spark</groupId>
    <artifactId>spark-core_2.11</artifactId>
    <version>2.4.1</version>
</dependency>
<dependency>
    <groupId>org.apache.spark</groupId>
    <artifactId>spark-sql_2.11</artifactId>
    <version>2.4.1</version>
</dependency>

Running Dockerized Demo Application

YuritaSampleApp directory in the Yurita project root path contains a standalone scala project for you to play around with. Run the demo through Docker inside YuritaSampleApp directory as shown below.

Build Docker Image

foo@bar:~/YuritaSampleApp$ docker build -f Dockerfile -t yuritademo .

Run Docker Container

foo@bar:~/YuritaSampleApp$ docker run -p 8080:8080 -t yuritademo

Writing Your First App

Create SparkSession with your own configurations

val appName = "AnomalyDetectionAPI"
val sparkConf = new SparkConf().setAppName(appName).setMaster("local[*]")
val spark = SparkSession
    .builder()
    .config(sparkConf)
    .getOrCreate()

Create dataframe of your data points/attributes with what time interval they occur on

//sample window timestamp
val window1 = (dateFormat.parse("2011-01-18 01:00:00.0"), dateFormat.parse("2011-01-18 01:00:10.0"))
val inputDF: DataFrame = Seq(
    Person("Ned", "Stark", 40, 40.6, "M", Array(5.5), getTimestamp(window1)),
    Person("Arya", "Stark", 9, 40.1, "F", Array(5.6), getTimestamp(window2)),
    Person("Sansa", "Stark", 13, 46.3, "F", Array(5.6), getTimestamp(window3)),
    Person("Jon Snow", "Stark", 17, 11.4, "M", Array(12.4), getTimestamp(window1),
    ...
).toDF()

Create a data pipe that will perform specified stastical methods on set columns of dataframe within the window size.

val categoricalPipe = PipelineBuilder()
    .onColumns(Seq("surname", "gender"))
    .setWindowing(Window.fixed("1 hour"))
    .setWindowReferencing(windowRef)
    .buildCategoricalModel(
    Functions.Categorical.avgRef,
    Functions.Categorical.entropy,
    Functions.statResultThreshold(3.0))

Combine multiple pipelines

val workload = AnomalyWorkload.builder()
    .addAllPipelines(categoricalPipe)
    .addPartitioner("surname")
    .buildWithWatermark("timestamp", "2 hours")

Dataset extended api

df.detectAnomalies(workload).map(_.toString).foreach(println(_))

Full demo application code can be viewed in our YuritaSampleApp project.

Contributing to Yurita

Thank you very much for contributing to Yurita. Please read the contribution guidelines for the process.

License

Yurita is licensed under the Apache License, v2.0

About

Anomaly detection framework @ PayPal

Resources

License

Code of conduct

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published