Online customized selection from multiple configurable standard models

Abstract

In electronic catalogs, commodities like computers and electronic equipment are specified as standard models even though a variety of possible alternative specifications can exist as a combination of selected options; therefore, customized configurations are essential to support various customers with individual needs. Most popular form of web support is displaying the standard models and matching the requirement by a similarity rule. Customers may modify the specification with the selected standard models.

However, the goals of selected factors may be incompatible with each other, and the modification with the most similar model may not guarantee the minimum price. To assist in the selection of the standard model and its modification process, we developed an algorithm named SMSM (Standard Model Selection and Modification). The first phase evaluates the standard models by the similarity measure as most Case-Based Reasoning (CBR) approaches do. The second phase modifies the specification of a standard model by the Constraint Satisfaction Problem approach, maintaining the compatibility. The CSP for the algorithms adopts concurrent local propagation and pruning owing to the nature of standard models. The second phase is repeated with the next most similar standard model and compares its configured specification with the previous ones until a local optimal specification is found.

We have applied the SMSM approach for the configuration of Dell’s PC products with 42 standard models with 25 variables. The approach was tested in 76 experimental cases. We discovered that 37 out of 76 cases could not find localoptimal configuration with the most similar standard models and that the localoptimal solutions reduce the cost by an average of 4.87%.