Based on reinforcement learning and fuzzy clustering theory, this paper proposes an algorithm OFC for the focused web crawler. We combine the naive Bayes classifiers with the fuzzy center-averaged clustering method to calculate the fuzzy memberships that are used to solve the valLie function mapping the hyperlinks to the future discounted rewards. Online estimation of the immediate action reward and classification of the newly crawled web pages incrementally enhance the crawling performance. We give several experiments to test the performance of algorithm OFC.