Modeling personal preferences on commodities by behavior log analysis with ubiquitous sensing

Abstract

Consumers may take some specific behavior preference or favorite items to get more information, such as the material and the price, in shopping. We have been developing a smart room to estimate their preference and favorite items through observation using ubiquitous sensors, such as RFID and Web cameras. We assumed the decision decision-making process in shopping as AIDMA rule, and detected specific behavior, which are "See", "Touch" and "Take", to estimate user's interest. We found that we can classify consumers by their behavior patterns of the times and duration of the behaviors. In our experiment we have tested twenty-eight subjects on twenty-four T-shirts. In the experiment, we got better precision ratio for each subjects on estimating preference and favorite items by discriminate analysis on his or her behavior log, and behavior patterns classification above. © 2009 Springer Berlin Heidelberg.