Systems and methods for classifying imbalanced data

Abstract

A transaction classification system obtains a data set including first data associated with a first characteristic and second data associated with a second characteristic. In response to obtaining the data set, the system uses a classification model to generate a classification by classifying the first data into majority data and the second data into minority data. From the classification and using the classification model, the system determines a cost. The system modifies the classification model based on this cost to generate an updated classification model. The system uses the updated classification model to re-classify a subset of the first data into the minority data and a subset of the second data into the majority data.

1 . (canceled) 
     
     
         2 . A computer-implemented method, comprising:
 receiving a new data set, wherein the new data set is generated using an existing classification model, and wherein the existing classification model was previously trained using an imbalanced data set and a modified loss function to satisfy one or more criteria;   processing the new data set using the existing classification model and the modified loss function to generate one or more quantifiable losses, wherein the one or more quantifiable losses correspond to a misclassification of one or more data points from the new data set;   determining that the one or more quantifiable losses do not satisfy the one or more criteria;   iteratively updating one or more model coefficients associated with the existing classification model until an updated classification model is obtained that produces a new quantifiable loss that satisfies the one or more criteria, wherein the one or more model coefficients are iteratively updated according to the new data set and the modified loss function; and   providing the updated classification model for use in processing incoming data, wherein the updated classification model is provided as a result of the updated classification model satisfying the one or more criteria.   
     
     
         3 . The computer-implemented method of  claim 2 , wherein the new data set is received with a new loss threshold, and wherein the new loss threshold is used to modify the one or more criteria. 
     
     
         4 . The computer-implemented method of  claim 2 , wherein the one or more quantifiable losses correspond to overall costs resulting from false positive classifications and false negative classifications from the new data set generated using the existing classification model. 
     
     
         5 . The computer-implemented method of  claim 2 , wherein processing the new data set includes:
 dividing the new data set into one or more data subsets; and   performing a set of evaluations of the existing classification model using the one or more data subsets, wherein the set of evaluations are performed to determine the one or more quantifiable losses.   
     
     
         6 . The computer-implemented method of  claim 2 , wherein the one or more criteria include a requirement whereby the one or more quantifiable losses are not to exceed resulting losses derived from processing of the new data set using one or more other machine learning models. 
     
     
         7 . The computer-implemented method of  claim 2 , wherein the one or more model coefficients are iteratively updated using gradient descent to generate new cutoff values usable to classify data points associated with the new data set. 
     
     
         8 . The computer-implemented method of  claim 2 , wherein the new data set corresponds to approved credit applications, and wherein the approved credit applications are classified as being authentic or as being fraudulent. 
     
     
         9 . A system, comprising:
 one or more processors; and   memory storing thereon instructions that, as a result of being executed by the one or more processors, cause the system to:
 receive a new data set, wherein the new data set is generated using an existing classification model, and wherein the existing classification model was previously trained using an imbalanced data set and a modified loss function to satisfy one or more criteria; 
 process the new data set using the existing classification model and the modified loss function to generate one or more quantifiable losses, wherein the one or more quantifiable losses correspond to a misclassification of one or more data points from the new data set; 
 determine that the one or more quantifiable losses do not satisfy the one or more criteria; 
 iteratively update one or more model coefficients associated with the existing classification model until an updated classification model is obtained that produces a new quantifiable loss that satisfies the one or more criteria, wherein the one or more model coefficients are iteratively updated according to the new data set and the modified loss function; and 
 provide the updated classification model for use in processing incoming data, wherein the updated classification model is provided as a result of the updated classification model satisfying the one or more criteria. 
   
     
     
         10 . The system of  claim 9 , wherein the new data set is received with a new loss threshold, and wherein the new loss threshold is used to modify the one or more criteria. 
     
     
         11 . The system of  claim 9 , wherein the one or more quantifiable losses correspond to overall costs resulting from false positive classifications and false negative classifications from the new data set generated using the existing classification model. 
     
     
         12 . The system of  claim 9 , wherein the instructions that cause the system to process the new data set further cause the system to:
 divide the new data set into one or more data subsets; and   perform a set of evaluations of the existing classification model using the one or more data subsets, wherein the set of evaluations are performed to determine the one or more quantifiable losses.   
     
     
         13 . The system of  claim 9 , wherein the one or more criteria include a requirement whereby the one or more quantifiable losses are not to exceed resulting losses derived from processing of the new data set using one or more other machine learning models. 
     
     
         14 . The system of  claim 9 , wherein the one or more model coefficients are iteratively updated using gradient descent to generate new cutoff values usable to classify data points associated with the new data set. 
     
     
         15 . The system of  claim 9 , wherein the new data set corresponds to approved credit applications, and wherein the approved credit applications are classified as being authentic or as being fraudulent. 
     
     
         16 . A non-transitory, computer-readable storage medium storing thereon executable instructions that, as a result of being executed by one or more processors of a computer system, cause the computer system to:
 receive a new data set, wherein the new data set is generated using an existing classification model, and wherein the existing classification model was previously trained using an imbalanced data set and a modified loss function to satisfy one or more criteria;   process the new data set using the existing classification model and the modified loss function to generate one or more quantifiable losses, wherein the one or more quantifiable losses correspond to a misclassification of one or more data points from the new data set;   determine that the one or more quantifiable losses do not satisfy the one or more criteria;   iteratively update one or more model coefficients associated with the existing classification model until an updated classification model is obtained that produces a new quantifiable loss that satisfies the one or more criteria, wherein the one or more model coefficients are iteratively updated according to the new data set and the modified loss function; and   provide the updated classification model for use in processing incoming data, wherein the updated classification model is provided as a result of the updated classification model satisfying the one or more criteria.   
     
     
         17 . The non-transitory, computer-readable storage medium of  claim 16 , wherein the new data set is received with a new loss threshold, and wherein the new loss threshold is used to modify the one or more criteria. 
     
     
         18 . The non-transitory, computer-readable storage medium of  claim 16 , wherein the one or more quantifiable losses correspond to overall costs resulting from false positive classifications and false negative classifications from the new data set generated using the existing classification model. 
     
     
         19 . The non-transitory, computer-readable storage medium of  claim 16 , wherein the executable instructions that cause the computer system to process the new data set further cause the computer system to:
 divide the new data set into one or more data subsets; and   perform a set of evaluations of the existing classification model using the one or more data subsets, wherein the set of evaluations are performed to determine the one or more quantifiable losses.   
     
     
         20 . The non-transitory, computer-readable storage medium of  claim 16 , wherein the one or more criteria include a requirement whereby the one or more quantifiable losses are not to exceed resulting losses derived from processing of the new data set using one or more other machine learning models. 
     
     
         21 . The non-transitory, computer-readable storage medium of  claim 16 , wherein the one or more model coefficients are iteratively updated using gradient descent to generate new cutoff values usable to classify data points associated with the new data set. 
     
     
         22 . The non-transitory, computer-readable storage medium of  claim 16 ,
 wherein the new data set corresponds to approved credit applications, and wherein the approved credit applications are classified as being authentic or as being fraudulent.