PORSCHE: A physical objects recommender system for cell phone users
Hideyuki Kawashima, Takaki Matsushita, Satoru Satake, Michita Imai, Yasuhide Shinagawa, Yuichiro Anzai · 2006
In web applications, recommendation algorithms such like collaborative flltering techniques are widely accepted for web users. To apply the algorithms for physical world objects such as books, three problems arise. The flrst problem is, how to input user ratings for physical objects ? The second problem is, how to measure user interests for physical objects ? The third problem is, how to consider the relationship between physical objects and users based on their locations ? In order to solve the problems, this paper proposes a physical objects recommender system for cell phone users, PORSCHE. As for the flrst problem, PORSCHE estimates the user ratings using user behaviors for physical objects. As for the second problem, PORSCHE changes the user ratings using both user behaviors and elapsed time. As for the third problem, PORSCHE adjusts the user ratings by continually monitoring the distance between physical objects and users. The result of experiments simulating a bookstore shows that PORSCHE detects user interests accurately and it also recommends proper books for the user.