Anomaly detection for an e-commerce pricing system

Abstract

This application relates to apparatus and methods for identifying anomalies within data, such as pricing data. In some examples, a computing device receives data updates and selects a machine learning model to apply to the data update. The computing device may train the machine learning model with features generated based on historical purchase order data. An anomaly score is generated based on application of the machine learning model. Based on the anomaly score, the data update is either allowed, or denied. In some examples, the computing device re-trains the machine learning model with detected anomalies. In some embodiments, the computing device prioritizes detected anomalies for further investigation. In some embodiments, the computing device identifies the cause of the anomalies by identifying at least one feature that is causing the anomaly.

What is claimed is: 
     
         1 . A system comprising:
 a computing device configured to:
 determine that a predetermined amount of time has elapsed; 
 determine that a value has been updated; 
 determine a machine learning model to apply to the value based on a type of the value that has been updated; 
 generate an anomaly score based on application of the machine learning model to the value; 
 determine that the value is an anomaly based on the anomaly score; and 
 transmit anomaly data identifying the value. 
   
     
     
         2 . The system of  claim 1 , where the computing device is configured to:
 obtain a plurality of features that include a cost feature;   generate a feature score for each of the plurality of features based on the plurality of features and the cost feature;   train the machine learning model based on the generated feature scores.   
     
     
         3 . The system of  claim 2 , wherein the computing device is configured to determine at least a portion of the plurality of features based on a log function of each respective feature of the plurality of features and the cost feature. 
     
     
         4 . The system of  claim 2 , wherein the plurality of features comprise at least one of a price feature, a binary feature, a categorical feature, and a hierarchical feature. 
     
     
         5 . The system of  claim 2 , wherein the plurality of features are labelled. 
     
     
         6 . The system of  claim 1 , wherein the computing device is configured to train the machine learning model when the predetermined amount of time has elapsed. 
     
     
         7 . The system of  claim 1 , wherein determining that the value is an anomaly comprises determining that the anomaly score is beyond a predetermined amount. 
     
     
         8 . The system of  claim 1 , wherein the computing device is configured to:
 store anomaly data identifying the anomaly to a database; and   re-train the machine learning model based on the stored anomaly data.   
     
     
         9 . The system of  claim 1 , wherein generating the anomaly score for the value is based on a mean of the value. 
     
     
         10 . The system of  claim 1 , wherein the machine learning model is a supervised machine learning model. 
     
     
         11 . The system of  claim 1 , wherein the value is at least one of a price and a cost of an item. 
     
     
         12 . The system of  claim 1 , wherein the computing device is configured to:
 generate a block update signal identifying that the value is an anomaly; and   transmit the block update signal to a pricing system.   
     
     
         13 . The system of  claim 12 , wherein the computing device is configured to:
 determine an impact score for the anomaly based on at least one of an estimated profit loss and a forgone revenue;   transmit the block update signal to a pricing system when the determined impact score is beyond a threshold.   
     
     
         14 . A method comprising:
 determining that a predetermined amount of time has elapsed;   determining that a value has been updated;   determining a machine learning model to apply to the value based on a type of the value that has been updated;   generating an anomaly score based on application of the machine learning model to the value;   determining that the value is an anomaly based on the anomaly score; and   transmitting anomaly data identifying the value.   
     
     
         15 . The method of  claim 14  further comprising:
 obtaining a plurality of features that include a cost feature; 
 generating a feature score for each of the plurality of features based on the plurality of features and the cost feature; 
 training the machine learning model based on the generated feature scores. 
 
     
     
         16 . The method of  claim 15  further comprising determining at least a portion of the plurality of features based on a log function of each respective feature of the plurality of features and the cost feature. 
     
     
         17 . The method of  claim 11  further comprising training the machine learning model when the predetermined amount of time has elapsed. 
     
     
         18 . The method of  claim 11  wherein the method comprises:
 generating a block update signal identifying that the value is an anomaly; and 
 transmitting the block update signal to a pricing system. 
 
     
     
         19 . A non-transitory computer readable medium having instructions stored thereon, wherein the instructions, when executed by at least one processor, cause a device to perform operations comprising:
 determining that a predetermined amount of time has elapsed;   determining that a value has been updated;   determining a machine learning model to apply to the value based on a type of the value that has been updated;   generating an anomaly score based on application of the machine learning model to the value;   determining that the value is an anomaly based on the anomaly score; and   transmitting anomaly data identifying the value.   
     
     
         20 . The non-transitory computer readable medium of  claim 17  further comprising instructions stored thereon that, when executed by at least one processor, further cause the device to perform operations comprising:
 obtaining a plurality of features that include a cost feature; 
 generating a feature score for each of the plurality of features based on the plurality of features and the cost feature; 
 training the machine learning model based on the generated feature scores.