Source code and application accompanying the online inferencing blog
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
src/main
README.md
build.gradle
settings.gradle

README.md

On Line Predictions in Kafka Streams

This Kafka Streams application demonstrates using embedded ML library Apache Mahout to perform OnlineLogisticRegression of flight data from the Bureau of Transportation Statistics

Specifically this application aims to do two things:

  1. Demonstrate the ability to perform online inferencing by joining a KStream with a GlobalKTable (airport id is the key) containing coefficients/model that can be used to predict the if a flight will arrive on time or not, by making an prediction with the flight data in the record.

  2. Update the model by a separate stream (Processor API) that collects flight data and when enough data is collected retrain a model and publish the updated coefficients to the Kafka topic backing the GlobalKTable, ensuring up to date predictions and keeping the model up to date in a streaming manner and hopefully improve our

Initially we'll observe a poor prediction rate, around 50%, basically a coin flip. But as we collect more data we are able to build a better model and publish the new updated model to the GlobalKTable, resulting in much better prediction rates somewhere between 80-90%.

Again the point of this application is not about machine learning algorithms per-se or how to build better machine-learning models, but that we can leverage the GlobalKTable to publish and updated model/coefficients and improve our on-line inferencing in steaming manner without having to do a batch job.

This project uses Gradle and after cloning/downloading it is recommended to first run the gradle command.

It is assumed that a Kafka instance already installed and running.

To run this application

  1. Create the following topics: onlineRegression-by-airport, raw-airline-data, ml-data-input, predictions .
  2. To build the initial model and populate the GlobalKTable run ./gradlew populateGlobalKTable from terminal window.
  3. Then start the KStreamsOnLinePredictions application with ./gradlew runOnlinePredictions
  4. From a separate terminal window start the data feed of flight data that the application will make predictions for and additionally will be used to create new models and update the GlobalKTable: run ./gradlew runDataFeed