Calibrated web personalization with adaptive recurrent computing

In recent times, large-scale data transmission and processing have become possible along with an increase in the processing and memory capacities of computer systems. With the advent of smart device technology, computing environments are being developed to support 'interaction and feedback' that is specific to each customer's individual behavior. By acquiring a user's known information and monitoring his interests by following his online behavior, it has become possible to use his changing interests as triggers to learn and make more appropriate recommendations. In an online trading or e-commerce setting, multiple items are often purchased at the same time, which makes it different from the problem of determining the degree of preference for a single item at a time, as in the case of a movie recommendation. This method adjusts recommendations dynamically over the course of browsing for other products by a user, taking into account how the degree of preference for one product may affect those for others, when trying to predict the degree of preference for the next item. In this paper, a product recommendation method is proposed that dynamically understands customer needs and considers the degree to which each product itself is preferred (degree of preference). Based on this evaluation, it decides whether or not to intervene in a customer's perception of their individual product preferences, resulting in a recommendation method that can adapt to the customer's needs to a high degree. Further, it is able to make such effective recommendations in the time period between a customer's search and his decision to purchase.