BAR-ILAN UNIVERSITY Attaining Fast and Successful Search in E-Commerce Environments

Current online stores suffer from the problem of information overload. There are too many items to offer, and customers find themselves lost due to the vast selection. As opposed to traditional stores, there is no or only less guidance that helps the customers as they search. This thesis proposes a new approach for searching in online stores. This approach is based on algorithms commonly used today in information filtering systems, but are hardly used in search engines, such as the ones proposed in online stores. We employ this approach for both keyword search and browse search. An implementation of the approach is presented. GMSIM is a simulated online store, based on the domain of GM's Auto Spare Parts. Several search guide algorithms are described and compared using experiments. The algorithms use the profile and history of the customer, her demographics and profile neighbors and her history neighbors in order to help in the search process. The experiments' results show that the suggested algorithms might indeed be usable and applicable in the domain of online stores.