Systems and methods for analyzing documents using machine learning techniques

Abstract

Systems and methods for activity risk management are disclosed. A system for activity risk management may include a memory storing instructions and at least one processor configured to execute instructions to perform operations including: classifying document data by identifying at least one marker in the document data, the at least one marker being associated with a document type; selecting an extraction model based on the document type; extracting model input data from the classified document data using the extraction model; applying a machine learning model to the extracted model input data to score the document data, the machine learning model having been trained with document data of a same document type as the document type associated with the at least one marker; and generating, based on the applying, a favorability output based on an amount of risk associated with the document data.

1 - 20 . (canceled) 
     
     
         21 . A system for risk prediction, the system comprising:
 at least one processor; and   a non-transitory computer-readable medium containing instructions that, when executed by the at least one processor, cause the at least one processor to perform operations comprising:
 classifying document data by identifying at least one marker in the document data, the at least one marker being associated with a document type; 
 selecting an extraction model based on the document type; 
 extracting model input data from the classified document data using the extraction model; 
 applying a machine learning model to the extracted model input data to score the document data, the machine learning model having been trained with document data of a same document type as the document type associated with the at least one marker; and 
 generating, based on the application of the machine learning model to the extract model input data, a favorability output based on an amount of risk associated with the document data. 
   
     
     
         22 . The system of  claim 21 , wherein classifying the document data comprises performing an optical character recognition (OCR) technique to a document to create machine-readable text. 
     
     
         23 . The system of  claim 21 , wherein classifying the document data includes using a random forest classifier. 
     
     
         24 . The system of  claim 21 , wherein the at least one marker comprises one or more of a word, a phrase, a frequency of text, a position of text relative to a document, a position of text relative to other text in a document, a sentence, a number, or a pictographic identifier. 
     
     
         25 . The system of  claim 21 , wherein the at least one marker is associated with document type by a machine learning model trained with user-created mappings between markers and document types. 
     
     
         26 . The system of  claim 21 , the operations further comprising determining that the machine learning model has insufficient model input data to provide a model output of a threshold confidence. 
     
     
         27 . The system of  claim 26 , the operations further comprising notifying a user to provide additional model input data. 
     
     
         28 . The system of  claim 26 , the operations further comprising prompting a device to re-capture document data. 
     
     
         29 . The system of  claim 21 , wherein the favorability output comprises an amount of risk associated with a transaction or individual associated with the document data. 
     
     
         30 . The system of  claim 21 , wherein the machine learning model is trained using input data from an entity other than an entity from which the document data was accessed. 
     
     
         31 . The system of  claim 21 , the operations further comprising applying a prediction model to predict a change in model input data that will improve the favorability output. 
     
     
         32 . The system of  claim 31 , the operations further comprising training the prediction model to learn, through an iterative feedback loop of model inputs, combinations of individual traits and transaction parameters correlated with improved favorability outputs. 
     
     
         33 . The system of  claim 21 , the operations further comprising modifying at least one model parameter based on the favorability output. 
     
     
         34 . The system of  claim 21 , the operations further comprising modifying at least one model parameter based on a user input. 
     
     
         35 . The system of  claim 21 , the operations further comprising updating the machine learning model based on inputs received from multiple systems. 
     
     
         36 . A method for risk prediction, comprising:
 classifying document data by identifying at least one marker in the document data, the at least one marker being associated with a document type;   selecting an extraction model based on the document type;   extracting model input data from the classified document data using the extraction model;   applying a machine learning model to the extracted model input data to score the document data, the machine learning model having been trained with document data of a same document type as the document type associated with the at least one marker; and   generating, based on the application of the machine learning model to the extract model input data, a favorability output based on an amount of risk associated with the document data.   
     
     
         37 . The method of  claim 36 , wherein classifying the document data comprises performing an optical character recognition (OCR) technique to a document to create machine-readable text. 
     
     
         38 . The method of  claim 36 , wherein the document data is classified by a random forest classifier. 
     
     
         39 . The method of  claim 36 , wherein the at least one marker comprises one or more of a word, a phrase, a frequency of text, a position of text relative to a document, a position of text relative to other text in a document, a sentence, a number, or a pictographic identifier. 
     
     
         40 . The method of  claim 36 , wherein the machine learning model is trained using input data from an entity other than an entity from which the document data was accessed.