Match outcome prediction

A project for the Distributed Artificial Intelligence and Intelligent Agents course (FTN, University of Novi Sad), built using Java EE technologies and Angular.

General

• An agent framework that enables building multi-agent systems
• For demonstration purposes, there have been implemented two types of agents
  • Chat application agents
    • UserAgent
    • UserHelperAgent
  • Match outcome prediction agents
    • CollectorAgent
    • PredictorAgent
    • MasterAgent

Overview of communication between match outcome prediction agents

• When a prediction method is called, a master and predictor agent are started locally and a collector agent is started on each node in a cluster
• Each collector agent replies to the predictor agent with the match data for selected teams
• When data from all nodes has been forwarded to the predictor agent, it predicts the match outcome and sends the result to the master agent
• Master agent displays the result via websocket

Setup

• Java Enterprise Application
  • Download Wildfly 11 application server
  • Replace existing standalone-full-ha.xml file with the one provided here
  • For the non-master cluster node you must provide master node name in the connection.properties file
  • Publish chat-ear.ear to /standalone/deployments folder
• Angular Application
  • Download Node.js (version 14.15.0 used for development)
  • Install Angular CLI (version 10.2.0 used for dvelopment)
  • Navigate to match-score-prediction-client project and type: npm install

Start the application

• Java Enterprise Application

  • Navigate to /bin folder of Wildfly and type:

  Windows

  standalone.bat -c standalone-full-ha.xml
  

  Linux

  ./standalone.sh -c standalone-full-ha.xml
  

• Angular Application

  • Navigate to match-score-prediction-client project and type:
  
  

  ng serve
  

  • To access system monitoring UI, type http://localhost:4200 in a browser
  • To access match outcome prediction UI, type http://localhost:4200/prediction in a browser