The invention relates to a dynamic rule base construction method based on maximum frequent patterns. The method comprises the following steps of inputting a Web log data set, carrying out data filtering on the Web log data set, converting a time format into time stamps, and mapping IPs and URLs into numbers; carrying out user identification and session identification and constructing a WASD; traversing each line of the WASD, filtering recent access behaviors of each user, forming a recent access sequence of the user, and forming a DWASD based on the recent access sequences of all users; constructing a Spark operation environment variable sc; calculating min_support; carrying out weight counting on prefixes and extracting frequent 1 item sets; traversing the frequent 1 item sets and recursively searching frequent item sets; calculating the maximum frequent item set, thereby obtaining a dynamic rule base based on the maximum frequent patterns. According to the method, the computing quantity is reduced, and the rule base with relatively high timeliness is rapidly and effectively constructed.