A novel OLDA evolving agent architecture. Issue 4 (4th July 2017)

• Record Type: Journal Article
• Authors: Sahu, Shashank; Agarwal, Rashi; Tyagi, Rajesh Kumar
• Is Part Of: Journal of statistics & management systems. Volume 20:Issue 4(2017)
• Page Start: 553
• Page End: 564
• Publication Date: 2017-07-04
• Subjects: Evolving Agent Architecture -- Agent-Oriented Software Development -- Software Agent -- Machine Learning -- Purchase Agent -- Regression Tree; 68T42; Statistics -- Periodicals; Mathematical models -- Periodicals; Mathematical models; Statistics; Periodicals; 519.5
• DOI: 10.1080/09720510.2017.1395175
• Languages: English
• ISSNs: 0972-0510

Abstract

Abstract: The objective of the paper is to define an architecture which can be used in designing of evolving agents. Proposed architecture consists of four modules and four repositories that enable evolving nature in the agent. The architecture can be used to design an agent that works on behalf of a user. The architecture has been experimented in MATLAB using regression tree learning method. Experiments have been performed on a sample of 1000 purchasing patterns. Two experiments have been carried out that show that an increment of 19.94% average success can be achieved by implementing evolving capability in the agent and 14.5 average evolution cycles are needed to achieve 100% success for the agent. Agent developed using proposed architecture is able to change itself without the involvement of human being and it supports agent-oriented software development.