INTELLIGENT AGENTS FOR INFORMATION FILTERING AND DATA MINING

Since the late '90s, intranets , extranets and the Internet have provided platforms for an explosion in the amount of data and information available to WWW users. The number of web-based sites continues to grow exponentially. The cost and availability of hardware, software and telecommunications currently continues to be at a level that user worldwide can afford. The ease of use and the availability of user -oriented web browsers, such as Netscape and Internet Explorer, have attracted many new computer users to the online world. These factors, among others, continue to create opportunities for the design and implementation of i-agents to assist users in doing complex computing tasks associated with the WWW.

There are three major approaches for building agents for the WWW. The first approach is to integrate i-agents into existing search engine programs. The agent follows predefined rules that it employs in its filtering decisions. Using this approach has several advantages.

The second approach is a rule-based approach. With this approach, an agent is given information about the application. A knowledge engineers is required to collect the required rules and knowledge for the agent.

The third approach is a training approach. In this approach the agent is trained to learn the preferences and actions of its user (Jensen, 2002).

This research aims to describe an intelligent agent that is able to perceive the world around it. That is, to recognize and evaluate events as they occur, determine the meaning of those events, and then take actions on behalf of the user(s). An event is a change of state within that agent's environment, such as when an email arrives and the agent is to filter the email (see Watson, 1997; Bigus & Bigus, 1998), or when new data or information becomes available in one of the many forms described earlier.

An i-agent must be able to process data. I-agents may have several processing strategies. They may be designed to use simple strategies (algorithms), or they could use complex reasoning and learning strategies to achieve their tasks. The success of i-agents depends on how much value they provide to their users (see Jensen, 2002; Lucas & Nissenbaum, 2000; Watson, 1997; Bigus & Bigus, 1998) and how easily they can be employed by their user(s).

I-agents in this research are used to retrieve data and information from the WWW. Technical issues of the implementation of the system using HTTP protocol are described. The Java programming language was used in this research to create an i-agent. The i-agent developed actively searches out desired data and information on the Web, and filters out unwanted data and information in delivering its results.