Smart Advisor and Search Optimizer: Web-based Applications of Fuzzy Rules, Intelligence Systems and Hierarchical Clustering fo

1 Dept. of Marketing, Robinson College of Business 2 Dept. of Computer Science, College of Arts and Sciences Abstract Smart Advisor and Search Optimizer are two intelligent Web agents helping users search Web information efficiently and make good decisions effectively. They are related to each other because Smart Advisor uses Search Optimizer as one component. Search Optimizer is a match maker between a user profile and search objectives, which can narrow down the search using a small number of screens of user input and well-defined categories. A new clustering algorithm for personal searches, group searches (peer groups), and all users searches is designed. In addition, this platform maintains member bookmarks with the ability for members to highlight some very frequently used bookmarks. The user has the ability to refine existing views or new searches to the extent possible and to bookmark the current search at any time. The search optimizer has been designed to allow flexibility and future scalability. Smart Advisor is a fuzzy expert agent with hierarchical fuzzy knowledge base. It has a natural interactive interview process for a user (i.e., a user can play a game with the Smart Advisor to gradually find out rational solutions), and produces advisory at different stages. For example, the college selection with relevant advice is a multi-feature-based complex process of assessment and decision for different domains (major, selectivity, college cluster, and individual college, etc.), hierarchical fuzzy assessment trees are designed based on hierarchical relationships among features so as to make a user do detailed assessment on a small group of relevant features on different level. A fuzzy assessment tree consists of many nodes including leaf nodes representing interview features provided by a user and high-level nodes representing abstract features generated by interview features. Such a bottom-up propagation calculation will eventually reach the root node of the fuzzy assessment tree, and generate a final possibility of college admission. The hierarchical advisory expert system is that hybrid experience of many experts is used to make a reliable decision, i.e., a student can virtually talk to many experienced experts in parallel to get robust assessment and stable advice. Importantly, it is a general framework that can be used in other applications such as eCommerce, expert assessment and self-evaluation systems.