Online fraud prevention using genetic algorithm solution
Online fraud prevention including receiving a rules set to detect fraud, mapping the rules set to a data set, mapping success data to members of the rules set, filtering the members of the rules set, and ordering members of the data set by giving priority to those members of the data set with a greater probability for being fraudulent based upon the success data of each member of the rule set in detecting fraud. Further, a receiver coupled to an application server to receive a rules set to detect fraud, and a server coupled to the application server, to map the rules set to a data set, and to map the success data to each members of the rules set. The server is used to order the various members of the data set by giving priority to those members of the data set with a greatest probability for being fraudulent.
1. A system, comprising:
one or more hardware processors; and
a memory storing computer-executable instructions, that in response to execution by the one or more hardware processors, causes the system to perform operations comprising
determining, based on applying a set of rules to account data, a set of target accounts that have previously been identified as fraudulent;
generating a plurality of rule trees based on subsets of the set of rules, each subset corresponding to a random combination of rules from the set of rules;
for each rule tree in the plurality of rule trees:
identifying a corresponding set of result accounts based on applying the rule tree to the account data;
determining an intersection between the corresponding set of result accounts and the set of target accounts; and
calculating a respective fitness score corresponding to the rule tree based on an amount of accounts included in the intersection between the corresponding set of result accounts and the set of target accounts; and
generating a new set of rules based on the calculated respective fitness scores.
2. The system of claim 1 , wherein the operations further comprise:
ranking each rule tree in the plurality of rule trees based on the calculated respective fitness scores.
3. The system of claim 1 , wherein the operations further comprise:
determining one or more rule trees from the plurality of rule trees, wherein the respective fitness score of each of the one or more rule trees satisfies a score threshold.
4. The system of claim 3 , wherein the generating the new set of rules further comprises:
performing a genetic programming mutation process with respect to a subset of the one or more rule trees.
5. The system of claim 4 , wherein the operations further comprise:
identifying a first rule tree from the one or more rule trees, the first rule tree having the highest respective fitness score of the one or more rule trees, wherein the subset of the one or more rule trees does not include the first rule tree.
6. The system of claim 3 , wherein the generating the new set of rules further comprises:
generating a set of cross-over rule trees by performing a genetic programming cross-over process with respect to a subset of the one or more rule trees.
7. The system of claim 6 , wherein the operations further comprise:
replacing a subset of the one or more rule trees with the set of cross-over rule trees.
8. The system of claim 7 , wherein a number of rule trees included in the subset of the one or more rule trees is equal to a number of rule trees included in the set of cross-over rule trees.
9. The system of claim 7 , wherein the respective fitness scores of the subset of the one or more rule trees is less than the respective fitness scores of the other rule trees of the one or more rule trees.
10. A method, comprising:
determining, by a computer system based on applying an initial set of rules to account data, a set of target accounts that have previously been identified as fraudulent;
generating a plurality of rule trees based on subsets of the initial set of rules, each subset corresponding to a random combination of rules from the initial set of rules;
determining respective fitness scores corresponding to each rule tree in the plurality of rule trees, wherein the determining the respective fitness scores comprises:
identifying a first set of result accounts based on applying a first rule tree of the plurality of rule trees to the account data; and
calculating a first fitness score corresponding to the first rule tree based on a comparison between accounts included in the first set of result accounts and accounts included in the set of target accounts; and
generating a new set of rules based on the respective fitness scores.
11. The method of claim 10 , further comprising:
determining a respective rank for each rule tree in the plurality of rule trees based on the respective fitness scores; and
identifying a set of rule trees from the plurality of rule trees, wherein the respective rank for each rule tree of the set of rule trees satisfies a rank threshold.
12. The method of claim 11 , further comprising:
outputting a set of result trees by performing a genetic programming process with respect to the identified set of rule trees, wherein the new set of rules are generated based on the set of result trees.
13. The method of claim 12 , wherein the genetic programming process includes at least one of a genetic programming mutation process or a genetic programming cross-over process.
14. The method of claim 12 , wherein the performing the genetic programming process with respect to the identified set of rule trees comprises:
identifying a second rule tree from the plurality of rule trees, the second rule tree having the highest respective fitness score among the other rule trees of the plurality of rule trees, wherein the second rule tree is excluded from the identified set of rule trees.
15. The method of claim 12 , wherein the generating the new set of rules comprises:
replacing a subset of the identified set of rule trees with the set of result trees, wherein the respective fitness score of each rule tree in the subset of the identified set of rule trees fails to satisfy as second rank threshold.
16. The method of claim 10 , further comprising:
identifying first account data corresponding to a first account based on transaction data associated with a transaction between two parties; and
determining whether the first account is fraudulent by applying the new set of rules to first account data.
17. A non-transitory computer readable medium storing computer-executable instructions that in response to execution by one or more hardware processors, causes a system to perform operations comprising:
determining, based on applying an initial set of rules to account data, a set of target accounts that have previously been identified as fraudulent;
generating a plurality of rule trees based on subsets of the initial set of rules, each subset corresponding to a random combination of rules from the initial set of rules;
determining respective fitness scores corresponding to each rule tree in the plurality of rule trees, wherein the determining the respective fitness scores comprises:
identifying a first set of result accounts based on applying a first rule tree of the plurality of rule trees to the account data; and
calculating a first fitness score corresponding to the first rule tree based on a comparison between accounts included in the first set of result accounts and accounts included in the set of target accounts; and
generating a new set of rules based on the respective fitness scores.
18. The non-transitory computer readable medium of claim 17 , wherein the operations further comprise:
iteratively executing a genetic programming algorithm on the plurality of rule trees until a termination condition is satisfied.
19. The non-transitory computer readable medium of claim 18 , wherein the termination condition includes a number of iterations of executing the genetic programming algorithm being met.
20. The non-transitory computer readable medium of claim 18 , wherein the genetic programming algorithm includes at least one of a genetic programming mutation or a genetic programming cross-over.