Conclusion and Discussion

In this chapter, the design and implementation of an intelligent product-brokering agent for m-commerce applications by using genetic algorithms and an evaluation function were presented. A simple prototype was implemented using Java, and the preliminary results obtained from the experiments have been promising.

One of the possible improvements to the current work is to distribute agents in a network instead of hosting entirely in the same computer. For m-commerce applications, this would mean that the agents could now be hosted by a commercial ISP. An m-commerce user would not want to spend huge sums of money on maintaining a wireless connection to the ISP or a phone company. Likewise, it is unrealistic for mobile devices, such as cellular phones and PDAs, to always be "online." Currently, some ISPs provide some forms of storage space for their subscribers to use to store files inside their servers. In extension to this, an ISP could now also offer to host the agents that were created by or for their subscribers with a reasonable fee. These agents could perform their tasks inside these servers and report back to their user when he or she reestablishes another connection with the ISP.

However, allowing agents to be distributed over the network will raise some issues, which the developer should look into before the system could be implemented. Because the agents are distributed, some form of communication protocol and ACL (Agent Communication Language) will have to be designed and incorporated into the system.