Data management using topic modeling

Abstract

Systems and methods for data management using machine learning and artificial intelligence techniques related to topic modeling on text comments are described. The text comments may correspond to a particular transaction conducted by a user. Machine learning text analysis is performed on the text comment to determine one or more topics associated with the text comment. The topic with the highest correlation to the text comment is assigned to the transaction claim. Based on the topic assigned to the transaction claim, various actions may be performed, including remedial actions on a user account. These techniques may be applicable to chargeback fraud, in some embodiments.

What is claimed: 
     
         1 . (canceled) 
     
     
         2 . A system, comprising:
 a non-transitory memory; and   one or more hardware processors coupled to the non-transitory memory and configured to read instructions from the non-transitory memory to cause the system to perform operations comprising: 
 receiving, via a network connection, a transaction claim associated with a user account from a client device; 
 obtaining unstructured text associated with the transaction claim, wherein the unstructured text comprises data received from a user associated with the user account; 
 analyzing the unstructured text using a machine learning model, wherein the machine learning model is configured to provide a topic correlation output based on analyzing a latent semantic structure of the unstructured text; 
 assigning, from a plurality of transaction claims topics, a particular transaction claim topic to the transaction claim based on the topic correlation output; 
 retrieving a transaction claim history associated with the user account, wherein the transaction claim history comprises a plurality of transaction claims, wherein each transaction claim in the transaction claim history has been classified into a corresponding transaction claim topic from the plurality of transaction claim topics; 
 determining, for the user account, a fraudulent risk based on a number of transaction claims classified into each transaction claim topic of the plurality of transaction claim topics; 
 applying one or more restrictions to the user account based at least in part on the fraudulent risk determined for the user account; and 
 authorizing or denying the transaction claim based on the one or more restrictions applied to the user account. 
   
     
     
         3 . The system of  claim 2 , wherein the machine learning model is a Latent Dirichlet allocation (LDA) statistical model. 
     
     
         4 . The system of  claim 3 , wherein the operations further comprise:
 identifying a tuning parameter for the LDA statistical model, the tuning parameter associated with allocation of topics to the unstructured text.   
     
     
         5 . The system of  claim 4 , wherein the tuning parameter indicates a value for a number of topics allocatable to the unstructured text. 
     
     
         6 . The system of  claim 4 , wherein identifying the tuning parameter for the LDA statistical model further comprises:
 determining a standard deviation of a restriction rate across topics within the unstructured text; and   selecting a value for the number of topics to be identified within the unstructured text based on the restriction rate.   
     
     
         7 . The system of  claim 4 , wherein the topic correlation output includes a plurality of transaction claim topics generated in proportion to topics occurring within the unstructured text. 
     
     
         8 . The system of  claim 7 , wherein the operations further comprise:
 determining a correlation score for each transaction claim topic of the plurality of transaction claim topics, the correlation scores for the plurality of transaction claim topics representing a probability distribution of the plurality of transaction claim topics for the unstructured text.   
     
     
         9 . A method, comprising:
 receiving, via network connection, a current transaction claim associated with a user account from a client device, the current transaction claim including unstructured text received from a user, the user account associated with a transaction claim history including prior transaction claims associated with one or more transaction claim topics of a plurality of transaction claim topics;   determining, using a machine learning model, a current transaction claim topic for the unstructured text, the current transaction claim topic determined based on a latent semantic structure detected by the machine learning model;   determining, for the user account, a fraudulent risk based on a number of transaction claims classified into each transaction claim topic of the plurality of transaction claim topics;   applying one or more restrictions to the user account based at least in part on the fraudulent risk determined for the user account; and   authorizing or denying the transaction claim based on the one or more restrictions applied to the user account.   
     
     
         10 . The method of  claim 9 , further comprising:
 identifying a tuning parameter for the machine learning model, the tuning parameter associated with allocation of topics to the unstructured text.   
     
     
         11 . The method of  claim 10 , wherein identifying the tuning parameter for the machine learning model further comprises:
 determining a standard deviation of a restriction rate across topics within the unstructured text; and   selecting a value for the number of topics to be identified within the unstructured text based on the restriction rate.   
     
     
         12 . The method of  claim 10 , wherein the tuning parameter indicates a maximum value for a number of topics allocatable to the unstructured text. 
     
     
         13 . The method of  claim 9 , wherein the current transaction claim topic is a first current transaction claim topic of a plurality of current transaction claim topics, the plurality of current transaction claim topics generated in proportion to topics occurring within the unstructured text and wherein the method further comprises:
 determining a correlation score for each transaction claim topic of the plurality of transaction claim topics, the correlation scores for the plurality of transaction claim topics representing a probability distribution of the plurality of transaction claim topics for the unstructured text.   
     
     
         14 . The method of  claim 13 , further comprising:
 assigning, from the plurality of current transaction claim topics, a particular transaction claim topic to the transaction claim based on the probability distribution.   
     
     
         15 . A non-transitory machine-readable medium having instructions stored thereon that are executed by a computer system to perform operations comprising:
 receiving, via network connection, a current transaction claim associated with a user account from a client device, the current transaction claim including unstructured text received from a user, the user account associated with a transaction claim history including prior transaction claims associated with one or more transaction claim topics of a plurality of transaction claim topics;   determining, using a machine learning model, a current transaction claim topic for the unstructured text, the current transaction claim topic determined based on a latent semantic structure detected by the machine learning model;   determining, for the user account, a fraudulent risk based on the current transaction claim, the transaction claim history, the current transaction claim topic, and the one or more transaction claim topics of the transaction claim history;   applying one or more restrictions to the user account based at least in part on the fraudulent risk determined for the user account; and   authorizing or denying the transaction claim based on the one or more restrictions applied to the user account.   
     
     
         16 . The non-transitory machine-readable medium of  claim 15 , wherein the operations further comprise:
 identifying a tuning parameter for the machine learning model, the tuning parameter associated with allocation of topics to the unstructured text.   
     
     
         17 . The non-transitory machine-readable medium of  claim 16 , wherein identifying the tuning parameter for the machine learning model further comprises:
 determining a standard deviation of a restriction rate across topics within the unstructured text; and   selecting a value for the number of topics to be identified within the unstructured text based on the restriction rate.   
     
     
         18 . The non-transitory machine-readable medium of  claim 16 , wherein the tuning parameter indicates a maximum value for a number of topics allocatable to the unstructured text. 
     
     
         19 . The non-transitory machine-readable medium of  claim 15 , wherein the current transaction claim topic is a first current transaction claim topic of a plurality of current transaction claim topics, the plurality of current transaction claim topics generated in proportion to topics occurring within the unstructured text and wherein the operations further comprise:
 determining a correlation score for each transaction claim topic of the plurality of transaction claim topics, the correlation scores for the plurality of transaction claim topics representing a probability distribution of the plurality of transaction claim topics for the unstructured text.   
     
     
         20 . The non-transitory machine-readable medium of  claim 19 , wherein the operations further comprise:
 assigning, from the plurality of current transaction claim topics, a particular transaction claim topic to the transaction claim based on the probability distribution.   
     
     
         21 . The non-transitory machine-readable medium of  claim 15 , wherein each claim transaction topic is associated with a restriction rate and a subset of account restrictions and wherein applying the one or more restrictions further comprises:
 selecting the one or more restrictions to apply to the user account based on the restriction rate and the subset of account restrictions associated with the current claim transaction topic.