Intelligent Personalization Agent for Product Brokering

Background

Although there is a tremendous increase in e-commerce activities, technology in enhancing consumers’ shopping experience remains primitive. Unlike real life department stores, there are no sales assistants to aid consumers in selecting the most appropriate product for users. Consumers are further confused by the large options and varieties of goods available. Thus, there is a need to provide on top of the provided filtering and search services (Bierwirth, 2000) an effective piece of software in the form of a product brokering agent to understand their needs and assist them in selecting suitable products.

A user’s interest in a particular product is often influenced by the product attributes that range from price to brand name. This research classifies attributes as accounted, unaccounted, and detected. The same attributes may also be classified as quantifiable or nonquantifiable attributes.

Accounted attributes are predefined attributes that the system is specially customised to handle. A system may be designed to capture the user’s choice in terms of price and brand name, making them accounted attributes. Unaccounted attributes have the opposite definition, and such attributes are not predefined in the ontology of the system (Guan & Zhu, 2004). The system does not understand whether an unaccounted attribute represents a model or a brand name. Such attributes merely appear in the product description field of the database. The system will attempt to detect the unaccounted attributes that affect the user’s preference and consider them as detected attributes. Thus, detected attributes are unaccounted attributes that are detected to be vital in affecting the user’s preference.

Quantifiable attributes contain specific numeric values (e.g., hard disk size) and thus their values are well defined. Nonquantifiable attributes, on the other hand, do not have any numeric values and their valuation may differ from user to user (e.g., brand name).