Applying Genetic Algorithms and Other Heuristic Methods to Handle PC Configuration Problems

Singapore is developing very fast as an Information Technology (IT) hub in which many people are keen to configure and build their own personal computers (PC). Like many real-life configuration problems, a well-designed PC configuration often represents a challenge in which given the wide diversity of hardware components, the ever-changing PC technology and the limited compatibility between some hardware components. we are interested to obtain an (sub-)optimal configuration for each specific usage restricted to a budget limit and other prefeeeef criteria. In this paper, we formally defined these PC configuration problems as discrete optimization problems. Then we proposed a systematic and flexible framework in which we can integrate any heuristic search method for solving these difficult real-world discrete optimization problems. A Possible advantage of our proposed framework is that users can flexibly add in or modify their specific requirements at any time. To demonstrate the feasibility of our proposal, we built a prototype of an intelligent Personal Computer Configuration Advisor available on the web to assist the general users in configuring their own PCs. Interestingly, our work opens up many new directions for future investigation including the improvement of our optimizer to handle more complicated users' requirements, and the possible uses of efficient learning algorithms such as the ID3 algorithm [2] to classify different user-configurations into useful examples to guide the search during optimization.