System and methods for message timing optimization

Abstract

Disclosed are systems, methods, and non-transitory computer-readable media for message timing optimization. A message timing optimization system determines optimized times to transmit messages to users based on historical transaction data. For example, the message timing optimization system may determine an optimal time to transmit a recommendation message to a user to perform a subsequent, such as repurchasing an item, refreshing a password, and the like. The message timing optimization system uses historical transaction data describing previous transactions performed by the user and/or other users to determine probability values indicating the likelihood that a user will perform a subsequent action at various time periods after the user performed an initial transaction.

What is claimed is: 
     
         1 . A method, performed by one or more processors, comprising:
 training, using historical transaction data, a machine learning model that identifies probability values for a subsequent transaction to occur at various periods of time, the training being based on features including geographic locations associated with accounts of users that have completed transactions;   identifying an account of a user associated with a completed transaction for an item, the completed transaction having occurred at a transaction time;   applying data associated with the completed transaction to the machine learning model to determine a first probability indicating a likelihood that the account will perform the subsequent transaction associated with the item during a first period of time after the transaction time and a second probability value indicating a likelihood that the account will perform the subsequent transaction associated with the item during a second period of time after the transaction time;   determining, based on the first probability value and the second probability value, that the subsequent transaction for the item is likely to be performed by the account during the first period of time after the transaction time;   determining a recommendation time to transmit a recommendation based on the first period of time and the transaction time;   transmitting a recommendation message to a contact identifier associated with the account at the recommendation time, the recommendation message providing a recommendation to perform the subsequent transaction for the item; and   retraining the machine learning model based on additional historical transaction data.   
     
     
         2 . The method of  claim 1 , wherein the training comprises generating vector representations based on historical transaction data associated with the account of the user, the historical transaction data associated with the account of the user being weighted higher for the training. 
     
     
         3 . The method of  claim 1 , wherein the training comprises generating vector representations based on historical transaction data associated with the item including a time of each initial and subsequent transaction. 
     
     
         4 . The method of  claim 1 , wherein the training comprises generating vector representations based on historical transaction data associated with multiple items within a category of the item, each category having a different category model. 
     
     
         5 . The method of  claim 1 , wherein determining the first probability value and the second probability value comprises:
 generating a feature vector based on data associated with the account; and   providing the feature vector as input into the machine learning model, the machine learning model providing the first probability value and the second probability value as output based on the feature vector.   
     
     
         6 . The method of  claim 1 , wherein the recommendation time to transmit the recommendation is a predetermined number of days prior to the first period of time. 
     
     
         7 . The method of  claim 1 , wherein the completed transaction is purchasing the item listed by an online service and the subsequent transaction is repurchasing the item. 
     
     
         8 . The method of  claim 1 , wherein the completed transaction is purchasing the item listed by an online service and the subsequent transaction is scheduling a service related to the item. 
     
     
         9 . The method of  claim 1 , wherein the subsequent transaction is a refreshing of a password. 
     
     
         10 . A system comprising:
 one or more computer processors; and   one or more computer-readable mediums storing instructions that, when executed by the one or more computer processors, cause the system to perform operations comprising:   training, using historical transaction data, a machine learning model that identifies probability values for a subsequent transaction to occur at various periods of time, the training being based on features including geographic locations associated with accounts of users that have completed transactions;   identifying an account of a user associated with a completed transaction for an item, the completed transaction having occurred at a transaction time;   applying data associated with the completed transaction to the machine learning model to determine a first probability indicating a likelihood that the account will perform the subsequent transaction associated with the item during a first period of time after the transaction time and a second probability value indicating a likelihood that the account will perform the subsequent transaction associated with the item during a second period of time after the transaction time;   determining, based on the first probability value and the second probability value, that the subsequent transaction for the item is likely to be performed by the account during the first period of time after the transaction time;   determining a recommendation time to transmit a recommendation based on the first period of time and the transaction time;   transmitting a recommendation message to a contact identifier associated with the account at the recommendation time, the recommendation message providing a recommendation to perform the subsequent transaction for the item; and   retraining the machine learning model based on additional historical transaction data.   
     
     
         11 . The system of  claim 10 , wherein the training comprises generating vector representations based on historical transaction data associated with the account of the user, the historical transaction data associated with the account of the user being weighted higher for the training. 
     
     
         12 . The system of  claim 10 , wherein the training comprises generating vector representations based on historical transaction data associated with the item including a time of each initial and subsequent transaction. 
     
     
         13 . The system of  claim 10 , wherein the training comprises generating vector representations based on historical transaction data associated with multiple items within a category of the item, each category having a different category model. 
     
     
         14 . The system of  claim 10 , wherein determining the first probability value and the second probability value comprises:
 generating a feature vector based on data associated with the account; and   providing the feature vector as input into the machine learning model, the machine learning model providing the first probability value and the second probability value as output based on the feature vector.   
     
     
         15 . The system of  claim 10 , wherein the recommendation time to transmit the recommendation is a predetermined number of days prior to the first period of time. 
     
     
         16 . The system of  claim 10 , wherein the completed transaction is purchasing the item listed by an online service and the subsequent transaction is repurchasing the item. 
     
     
         17 . The system of  claim 10 , wherein the completed transaction is purchasing the item listed by an online service and the subsequent transaction is scheduling a service related to the item. 
     
     
         18 . The system of  claim 10 , wherein the subsequent transaction is a refreshing of a password. 
     
     
         19 . A non-transitory computer-storage medium storing instructions that, when executed by one or more computer processors of one or more computing devices, cause the one or more computing devices to perform operations comprising:
 training, using historical transaction data, a machine learning model that identifies probability values for a subsequent transaction to occur at various periods of time, the training being based on features including geographic locations associated with accounts of users that have completed transactions;   identifying an account of a user associated with a completed transaction for an item, the completed transaction having occurred at a transaction time;   applying data associated with the completed transaction to the machine learning model to determine a first probability indicating a likelihood that the account will perform the subsequent transaction associated with the item during a first period of time after the transaction time and a second probability value indicating a likelihood that the account will perform the subsequent transaction associated with the item during a second period of time after the transaction time;   determining, based on the first probability value and the second probability value, that the subsequent transaction for the item is likely to be performed by the account during the first period of time after the transaction time;   determining a recommendation time to transmit a recommendation based on the first period of time and the transaction time;   transmitting a recommendation message to a contact identifier associated with the account at the recommendation time, the recommendation message providing a recommendation to perform the subsequent transaction for the item; and   retraining the machine learning model based on additional historical transaction data.   
     
     
         20 . The non-transitory computer-storage medium of  claim 19 , wherein determining the first probability value and the second probability value comprises:
 generating a feature vector based on data associated with the account; and   providing the feature vector as input into the machine learning model, the machine learning model providing the first probability value and the second probability value as output based on the feature vector.