Alternative goods recommender by customer´s context - example for business hotel room reservation

This paper presents an active recommender system for TPO (time, place, and occasion)-dependent goods. Basic premise under this research is that customer´s preference for selecting goods is embedded in his past purchase history. In order to recommend the alternative goods, our system (1) checks whether the current customer´s selection differs from the past choice or not, and (2) finds how to adjust current selection to his preference. The presented system adopts context metrization by statistic methods for the former function and memory-based reasoning (MBR) by personal probability distribution for the latter function. Applying the presented system to the business hotel room reservation service which includes 10,700 customers and 400,000 transaction records, this paper shows that the accuracy is higher than seventy percent.