Customers are increasingly demanding products and services that satisfy their specific needs. Mass customization is the mass production of individually tailored products that satisfy the specific needs of each customer. It involves the acquisition and satisfaction of customer requirements. Product customization is costly because each customer requires individualized attention. There is a need for intelligent software that identifies possible customer preferences and then assembles products based on the identified preferences. This study provides a framework and an algorithm for the interactive customization of products and services (Iona).; The framework consists of the following functions: acquisition, assessment, elimination, selection and explanation/description. The following information must be acquired: (1) absolute/preferred constraints (product component specifications considered most important), (2) categorical preferences, (3) stereotype (customer type), and (4) context (purchase situation). Assessment consists of the following: (1) deciding which choices violate the absolute/preferred constraints, the categorical constraints, and the constraints inherent in the product (binary constraints between the different parts of the product and unary constraints such as availability), and (2) calculating the multi-attribute utility (usefulness to the customer based on the attribute levels) of the remaining choices. Choices that violate the constraints are eliminated to arrive at a basic solution. A selection is made for each part of the product using a binary integer programming model after choice utilities are estimated at an acceptable level of certainty. Questions regarding stereotypes/contexts are asked of the user to refine the utility estimates of the choices. The query to generate is based on the user's possible membership in the stereotype/context and the probable improvement in the utility estimates. A choice is selected for each part of the product when the usefulness of asking a query is less than the cost of querying. When providing categorical preferences, descriptions of the discriminating attributes are given for users unfamiliar with the product. The final solution is explained emphasizing the attributes considered important based on the user model. The algorithm is applied to the domain of travel planning. A prototype (Travel Planner) demonstrates the feasibility of the framework and the algorithm.
展开▼