A novel ranking algorithm based on Reinforcement Learning

People are very interested in finding most related web pages when they are looking for some information about a topic in the web. This is the responsibility of the search engine to use effective ranking methods to find the most related web pages for a given query. In this paper, we propose a novel ranking method which is based on connectivity and also benefits from the Reinforcement Learning (RL) concepts. RL problems are structured around estimating value functions. In our algorithm, each web page is considered as a state and its score is as value function of the state. So, the key elements in our method are agent, value function and considered interstate transition rewards. Also we inspected a hybrid ranking algorithm which combined results of several basic ranking algorithms. The proposed algorithms are evaluated by using well known benchmark data sets and they are analyzed according to concerning criteria. Experimental results show that applying reinforcement learning method leads to considerable improvements and the hybrid algorithm can outperform all other basic ranking algorithms.