Enhanced validity modeling using machine-learning techniques
The present disclosure generally relates to a primary load management system configured to execute machine learning and artificial intelligence techniques to generate predictions of access-right requests that are or are likely to be invalid before the access-right requests are processed for assignment to users or user devices. The present disclosure relates to systems and methods that collect a data set representing characteristics of user devices as the user devices interact with various systems of the primary load management system and train a machine-learning model to predict invalid access-right requests using the collected data set. The collected data set may include a log line that represents each user device, and each log line may be labeled based on an invalidity evaluation. New access-right requests can be processed using the trained machine-learning model to determine whether or not to assign access rights in response to the access-right request.
1 . A method for detecting access right requests associated with a valid action or an invalid action by processing data streams, the method comprising:
collecting a feature set of one or more features characterizing a plurality of users, wherein each feature of the feature set corresponds to an attribute of a previous interaction between a user device and a primary load management system;
storing a structured data string, representing each user of the plurality of users, wherein the structured data string contains at least a subset of the feature set;
training one or more classifier models by the structured data string;
transmitting at least two data streams in real-time from an event data page to the primary load management system, wherein the at least two data streams include a plurality of data elements, wherein each data element of the plurality of data elements represents the access right requests for one or more access rights associated with a resource;
feeding the at least two data streams into a data joining system;
identifying a common join key in each data element of the plurality of data elements by the data joining system;
generating new data streams by joining the plurality of data elements from the at least two data streams using the identified common join key;
inputting the new data streams into one or more classifier models to classify the access right requests associated with the valid action or the invalid action;
generating by the one or more classifier models, a recommendation signal to deny or grant the access right requests associated with the resource in response to inputting the new data streams, wherein:
the granting of the access right requests indicates that the access right requests are associated with the valid action;
the denial of the access right requests indicates that the access right requests are associated with the invalid action;
generating by a decision generator, a decision for each of the access right requests by feeding the recommendation signal and the new data streams into the decision generator, wherein the decision indicates whether to deny or grant the access right requests; and
performing an action based on the generated decision, wherein the generated decision corresponds to a recommendation to:
grant the access right requests for the one or more access rights associated with the resource, performing the action includes enabling an assignment of the one or more access rights to the user device, or
deny the access right requests for the one or more access rights associated with the resource, performing the action includes blocking the assignment of the one or more access rights to the user device.
2 . The method according to claim 1 , wherein the invalid action corresponds to access right requests associated with fraudulent activity, including access right requests transmitted by a bot user or an unauthorized user.
3 . The method according to claim 1 , wherein the common join key is a unique identifier associated with each data element of the plurality of data elements.
4 . The method according to claim 1 , wherein the at least two data streams are live data streams.
5 . The method according to claim 1 , wherein the one or more classifier models comprises one or more machine learning models selected from, Naïve Bayes models, deep learning neural network models, logistic regression models, decision tree models, and support vector machines.
6 . The method according to claim 1 , wherein the primary load management system stores an access code associated with the access right requests.
7 . The method according to claim 1 , wherein the user device is granted access to the one or more access rights for a definite time period on granting the access right requests.
8 . The method according to claim 1 , wherein the recommendation signal represent the decision to deny the access right requests, as the access right requests are potentially being associated with the invalid action.
9 . The method according to claim 1 , wherein the recommendation signal represent the decision to grant the access right requests, as the access right requests are being associated with the valid action.
10 . The method according to claim 1 , wherein the decision generator is an assignment generator that weighs the recommendation signal received from each of the classifier model of the one or more classifier models to generate a final decision.
11 . The method according to claim 10 , wherein weighing of the one or more classifier models is predefined, and the decision to grant or deny the access right requests is based on a composite score or predetermined rule.
12 . The method according to claim 11 , wherein the access right requests are denied when the composite score is below a threshold.
13 . A system for detecting access right requests associated with a valid action or an invalid action by processing data streams, the system configured to:
collect a feature set of one or more features characterizing a plurality of users, wherein each feature of the feature set corresponds to an attribute of a previous interaction between a user device and a primary load management system;
store a structured data string, representing each user of the plurality of users, wherein the structured data string contains at least a subset of the feature set;
train one or more classifier models by the structured data string;
transmit at least two data streams in real-time from an event data page to the primary load management system, wherein the at least two data streams include a plurality of data elements, wherein each data element of the plurality of data elements represents the access right requests for one or more access rights associated with a resource;
feed the at least two data streams into a data joining system;
identify a common join key in each data element of the plurality of data elements by the data joining system;
generate new data streams by joining the plurality of data elements from the at least two data streams using the identified common join key;
input the new data streams into one or more classifier models to classify the access right requests associated with the valid action or the invalid action;
generate by the one or more classifier models, a recommendation signal to deny or grant the access right requests associated with the resource in response to inputting the new data streams, wherein:
the grant of the access right requests indicates that the access right requests are associated with the valid action;
the deny of the access right requests indicates that the access right requests are associated with the invalid action;
generate by a decision generator, a decision for each of the access right requests by feeding the recommendation signal and the new data streams into the decision generator, wherein the decision indicates whether to deny or grant the access right requests; and
perform an action based on the generated decision, wherein the generated decision corresponds to a recommendation to:
grant the access right requests for the one or more access rights associated with the resource, performing the action includes enabling an assignment of the one or more access rights to the user device, or
deny the access right requests for the one or more access rights associated with the resource, performing the action includes blocking the assignment of the one or more access rights to the user device.
14 . The system as recited in claim 13 , wherein the invalid action corresponds to access right requests associated with fraudulent activity, including access right requests transmitted by a bot user or an unauthorized user.
15 . The system as recited in claim 13 , wherein the common join key is a unique identifier associated with each data element of the plurality of data elements.
16 . The system as recited in claim 13 , wherein the at least two data streams are live data streams.
17 . The system as recited in claim 13 , wherein the one or more classifier models comprises one or more machine learning models selected from Naïve Bayes models, deep learning neural network models, logistic regression models, decision tree models, and support vector machines.
18 . The system as recited in claim 13 , wherein the primary load management system stores an access code associated with the access right requests.
19 . The system as recited in claim 13 , wherein the user device is granted access to the one or more access rights for a definite time periods on granting the access right requests.
20 . A computer-program product tangibly embodied in a non-transitory machine-readable storage medium, including instructions configured to cause a processing apparatus to perform operation for detecting access right requests associated with a valid action or an invalid action by processing data streams, including:
collecting a feature set of one or more features characterizing a plurality of users, wherein each feature of the feature set corresponds to an attribute of a previous interaction between a user device and a primary load management system;
storing a structured data string, representing each user of the plurality of users, wherein the structured data string contains at least a subset of the feature set;
training one or more classifier models by the structured data string;
transmitting at least two data streams in real-time from event data page to the primary load management system, wherein the at least two data streams include a plurality of data elements, wherein each data element of the plurality of data elements represents the access right requests for one or more access rights associated with a resource;
feeding the at least two data streams into a data joining system;
identifying a common join key in each data element of the plurality of data elements by the data joining system;
generating new data streams by joining the plurality of data elements from the at least two data streams using the identified common join key;
inputting the new data streams into one or more classifier models to classify the access right requests associated with the valid action or the invalid action;
generating by the one or more classifier models, a recommendation signal to deny or grant the access right requests associated with the resource in response to inputting the new data streams, wherein:
the granting of the access right requests indicates that the access right requests are associated with the valid action;
the denial of the access right requests indicates that the access right requests are associated with the invalid action;
generating by a decision generator, a decision for each of the access right requests by feeding the recommendation signal and the new data streams into the decision generator, wherein the decision indicates whether to deny or grant the access right requests; and
performing an action based on the generated decision, wherein the generated decision corresponds to a recommendation to:
grant the access right requests for the one or more access rights associated with the resource, performing the action includes enabling an assignment of the one or more access rights to the user device, or
deny the access right requests for the one or more access rights associated with the resource, performing the action includes blocking the assignment of the one or more access rights to the user device.