Method and system for scoring quality of traffic to network sites

A non-transitory computer readable medium containing a set of computer readable instructions for causing a computer to perform a method for identifying statistically significant rules from a global rule set to generate an environmental rule set in order to improve a scoring of a quality of traffic pertaining to advertisements displayed on a network site, the method comprising:

• automatically generating at least one rule in the global rule set by a traffic scoring facility, wherein the at least one rule identifies at least one environment feature that correlates with fraudulent traffic;
• generating the environmental rule set from the global rule set, by training a computer-based system for characterizing quality of agent interactions with the network site, wherein the global rule set includes a plurality of rules, each of the plurality of rules comprises an expression for evaluating one or more parameters associated with agent interactions with the network site during an agent session, wherein the generating the environmental rule set includes:
• receiving, from a server log, data describing a plurality of sessions with the network site, the data describing the one or more parameters associated with the plurality of sessions including an IP address and a domain name associated with the IP address, an agent ID, a record of a user input and a time the user input was made;
• identifying a particular agent session from the data describing the plurality of sessions, wherein the agent session is characterized by interactions with the network site by the agent ID;
• determining, by the computer-based system, agent interactions resulting in a desired agent action and agent interactions resulting in a non-desired agent interaction for each rule of the environmental rule set, wherein the desired agent action is selected from a group including a click, a purchase, a submission of a form, a bookmarking, a rollover event, or an impression;
• calculating, by the computer-based system, a correlation coefficient for each rule of the environmental rule set, the correlation coefficient for each rule being a numerator calculated by a percentage of a total number of agent actions in which the desired action occurred and a respective rule is satisfied minus a product of a percentage of total number of actions that are desired and a percentage of total number of agent actions in which the respective rule is satisfied, divided by a denominator calculated by a square root of a product of the percentage of total number of actions that are desired, the percentage of total number of agent actions in which the rule is satisfied, a percentage of total number of agent interactions in which a desired action does not occur, and a percentage of the total number of agent actions in which the rule is not satisfied;
• automatically identifying by the computer-based system, statistically significant rules of the environmental rule set whose calculated correlation coefficient exceeds a predetermined threshold;
• and
• assessing the quality of the one or more agent interactions based on a non-binary score which is generated by using the identified statistically significant rules.