/infinity

Prototype of forecast service that uses machine learning to deliver forecasts

Primary LanguageJavaApache License 2.0Apache-2.0

Infinity

Build Java Quarkus Camel Cassandra Kafka Spark Vue License

Infinity is a prototype of cloud-agnostic forecasting platform inspired by Amazon Forecast service.
Project was created as a part of the DataStax Hackathon aka ✨ASTRAKATHON✨ and won the first place.

Requirements

  • User should be able to upload dataset file
  • User should be able to publish events through an API
  • User should be able to view uploaded data
  • User should be able to start analysis (aggregations and predictions)
  • User should be able to view aggregations and predictions as tables and charts
  • System should store events permanently
  • System should store aggregations permanently
  • System should store predictions permanently
  • System should be horizontally scalable

How it works

Architecture

Architecture

Components

WebUI

Demo application to present results (infinity-rest).
User could upload CSV files with data for analysis and check results.
Implemented with Vue.js and Chart.js. See screenshots.

REST services

API to interact with the system (infinity-rest).
Implemented with Quarkus and Cassandra extension.

Processor

Data processing application (infinity-processor).
Includes three consumer groups to retrieve events from Kafka and store into Cassandra tables.
Implemented with Quarkus and Apache Camel.

Analytics

Data analytics application (infinity-analytics)
Aggregates events by SECOND, MINUTE, HOUR, DAY, MONTH and YEAR
and calculate AVG, MIN, MAX, MEAN, SUM, COUNT for event values.

Forecast values for aggregated values for all horizons.
Current version provides predictions with ARIMA algorithm for six steps.
Implemented with Spark and Apache Camel.

Kafka

Event store in CQRS architecture

Cassandra

Database for events, aggregations and predictions.
Tables:

  • EVENTS_BY_ID
  • EVENTS_BY_TIMESTAMP
  • EVENTS_BY_TIME
  • AGGREGATIONS
  • PREDICTIONS

Init

Init container to create Cassandra keyspace and tables (infinity-init)

Build and run

Requires Git, Docker and Docker Compose installed.

git clone git@github.com:mgubaidullin/infinity.git
docker-compose build
docker-compose up

Application is ready to use after following line in the log: infinity-init exited with code 0

Execute

User Interface

Open following link in browser http://localhost:8080

Upload data

  • Select file (quebec.csv) and click 'Upload' button

Data

  • Refresh page to review results (processing might take 5 seconds)

Data with values

  • Click 'Analyze' button to start aggregation and forecast
  • Go to Aggregations page to review aggregation results (analysis might take 20 seconds)

Aggregations

  • Go to Predictions page to review predictions results

Predictions

  • Go to Chart page to compare facts and forecast

Chart

Command line

Upload file with events

curl -i -X POST -H "Content-Type: multipart/form-data" -F "file=@quebec.csv" http://localhost:8080/file

Start analytics for special event group and type

curl -X POST "http://0.0.0.0:8080/analytic" -H  "accept: application/json" -H  "Content-Type: application/json" -d "{\"eventGroup\":\"Quebec\",\"eventType\":\"Trucks\"}"

Retrieve aggregations

curl -X GET "http://0.0.0.0:8080/analytic/aggregation/Quebec/Trucks/YEARS/2020" -H  "accept: application/json"

Retrieve predictions

curl -X GET "http://0.0.0.0:8080/analytic/prediction/Quebec/Trucks/ARIMA/YEARS/2025" -H  "accept: application/json"

Swagger

Swagger UI for API http://localhost:8080/swagger-ui