Synthetic clickstream testing using a neural network

Abstract

Disclosed herein are system, method, and computer program product embodiments for simulating users for testing websites using a trained model. Simulating users involves training a model to generate realistic synthetic clickstreams that emulate actual clickstreams. Synthetic clickstreams may include simulated mouse actions and keystrokes and the results of the testing may be used to improve the design and functioning of the tested website.

What is claimed: 
     
         1 . A computer-implemented method for generating synthetic user clickstreams by one or more computing devices of a trained synthetic stream generator, the method comprising:
 retrieving a stored clickstream that includes a user interaction including one of a keystroke or a mouse action;   classifying, by a first neural network of the trained synthetic stream generator, the user interaction from the stored clickstream into a classification;   translating, based on the classification, the user interaction into a feature that numerically represents the user interaction; and   generating, by a second neural network of the trained synthetic stream generator, a synthetic user clickstream based on the feature, wherein the synthetic user clickstream represents a predicted user interaction.   
     
     
         2 . The method of  claim 1 , wherein the mouse action includes at least one of a mouse movement, a mouse click, or a temporal relationship between the mouse action and a keystroke action. 
     
     
         3 . The method of  claim 1 , wherein the synthetic user clickstream comprises a predicted clickstream and wherein generating the synthetic user clickstream based on the feature further comprises:
 analyzing the feature over a time period;   detecting a pattern in the feature within the time period, wherein the pattern indicates a temporal relationship between a first feature and a second feature within the feature; and   forming the predicted clickstream based on the detected pattern.   
     
     
         4 . The method of  claim 3 , wherein the first feature is a mouse movement and the second feature is a mouse action. 
     
     
         5 . The method of  claim 1 , wherein based on detecting that the synthetic user clickstream is a synthetic clickstream, the method further comprising:
 providing an analysis response to the second neural network, wherein the analysis response indicates a result of comparing the synthetic user clickstream to an actual clickstream; and   generating, by the second neural network of the trained synthetic stream generator, an updated synthetic user clickstream based on the analysis response.   
     
     
         6 . The method of  claim 1 , the method further comprising:
 detecting a similarity between the synthetic user clickstream and an actual clickstream, wherein the similarity is based on a predetermined success rate in discriminating the synthetic user clickstream.   
     
     
         7 . The method of  claim 1 , wherein the first neural network is a convolutional recurrent neural network. 
     
     
         8 . The method of  claim 7 , wherein the convolutional recurrent neural network comprises a convolutional neural network and a recurrent neural network. 
     
     
         9 . The method of  claim 1 , wherein the second neural network is a generative adversarial network. 
     
     
         10 . The method of  claim 1 , wherein the stored clickstream is associated with a plurality of users, the method further comprising:
 clustering, by the one or more computing devices, the stored clickstream based on a user archetype to form at least a first cluster of the stored clickstream and a second cluster of the stored clickstream, wherein the synthetic user clickstream is further generated based on the first cluster of the stored clickstream; and   selecting the trained synthetic stream generator from a plurality of trained synthetic stream generators based on the user archetype.   
     
     
         11 . The method of  claim 10 , wherein the user archetype includes at least one of a user age, a user network speed, a user income, and a characteristic of a device associated with a user. 
     
     
         12 . The method of  claim 11 , wherein the characteristic of the device includes at least one of an operating system, a browser, and battery life. 
     
     
         13 . The method of  claim 10 , further comprising:
 generating, by the one or more computing devices, second synthetic user clickstream based on the second cluster of the stored clickstream and on the feature, wherein the second synthetic user clickstream represents a second predicted user interaction.   
     
     
         14 . A non-transitory computer-readable medium storing instructions, the instructions, when executed by a processor, cause the processor to perform operations comprising:
 retrieving a stored clickstream that includes a user interaction includes one of a keystroke or a mouse action;   classifying, by a first neural network of a trained synthetic stream generator, the user interaction from the stored clickstream into a classification;   translating, based on the classification, the user interaction into a feature that numerically represents the user interaction; and   generating, by a second neural network of the trained synthetic stream generator, a synthetic user clickstream based on the feature, wherein the synthetic user clickstream represents a predicted user interaction.   
     
     
         15 . The non-transitory computer-readable medium of  claim 14 , wherein the mouse action includes at least one of a mouse movement, a mouse click, or a temporal relationship between the mouse action and a keystroke action. 
     
     
         16 . The non-transitory computer-readable medium of  claim 14 , wherein the synthetic user clickstream comprises a predicted clickstream and wherein generating the synthetic user clickstream based on the feature further comprises:
 analyzing the feature over a time period;   detecting a pattern in the feature within the time period, wherein the pattern indicates a temporal relationship between a first feature and a second feature within the feature; and   forming the predicted clickstream based on the detected pattern.   
     
     
         17 . The non-transitory computer-readable medium of  claim 16 , wherein the first feature is a mouse movement and the second feature is a mouse action. 
     
     
         18 . The non-transitory computer-readable medium of  claim 14 , wherein the stored clickstream is associated with a plurality of users, the operations further comprising:
 clustering the stored clickstream based on a user archetype to form at least a first cluster of the stored clickstream and a second cluster of the stored clickstream, wherein the synthetic user clickstream is further generated based on the first cluster of the stored clickstream; and   selecting the trained synthetic stream generator from a plurality of trained synthetic stream generators based on the user archetype.   
     
     
         19 . The non-transitory computer-readable medium of  claim 18 , wherein the user archetype includes at least one of a user age, a user network speed, a user income, and a characteristic of a device associated with a user. 
     
     
         20 . An apparatus implementing a trained synthetic stream generator for testing a web site, comprising:
 a memory; and   a processor communicatively coupled to the memory and configured to:
 retrieve a stored clickstream that includes a user interaction includes one of a keystroke or a mouse action; 
 classify, by a first neural network of the trained synthetic stream generator, the user interaction from the stored clickstream into a classification; 
 translate, based on the classification, the user interaction into a feature that numerically represents the user interaction; and 
   generate, by a second neural network of the trained synthetic stream generator, a synthetic user clickstream based on the feature, wherein the synthetic user clickstream represents a predicted user interaction.