Website navigation recommendation based on reinforcement learning technique

Abstract

The explosive growth of the Internet has made information on the web large and complicated. If the structure of a website is not optimized, users could easily get lost and could not find the most important information at the first time. The adaptive website can present the information that users needed by analyzing the users’ behavior. However, visitors may have different needs at different times. Most of recommended methods are not considerate of dynamic or time-dependent needs. This paper presents a recommender system based on reinforcement learning. We assume that five parameters are on recommendation, which include clicks of the page, time that spent on viewing the page, paths to find the page, hierarchy of the page, and the rank of the page. With the help of reinforcement learning to adjust the weight of five parameters, we aim to reduce the paths that user needed to find the object page.