A computer program product comprising a non-transitory computer readable storage medium having instructions encoded thereon, the instructions for determining a real-time selection during a web browsing session, wherein the instructions, when executed by a processor, cause the processor to:
- receive data regarding prior data transmission transactions of a user, each prior data transmission transaction of the user corresponding to the user's usage of a particular data transmission rail, a data transmission rail being an architecture of how data is transmitted over a network;
- train a machine learning model using supervised learning based on the prior data transmission transactions to determine probabilities of the user completing web browsing sessions, each web browsing sessions being a transaction involving an order and completing a web browsing session comprising completing the transaction, the machine learning model trained from the prior data transmission transactions and outputting a probability of the user completing a particular web browsing session;
- receive, at an online system during the real-time web browsing session, information describing a real-time data transmission transaction, information identifying the user, and account information of a first transaction account for the user, wherein the first transaction account is associated with a first data transmission rail, the real-time data transmission transaction being in progress during the real-time web browsing session and being adapted to use the first data transmission rail to transmit data, wherein the real-time data transmission transaction includes data that is fixed at a checkout page of the web browsing session, and the first transaction account is managed by a first institution;
- apply, in real time during the real-time web browsing session, the machine learning model to the information describing the data that is fixed at the checkout page to identify an alternative transaction that is associated with a second data transmission rail, wherein the machine learning model predicts, in real time during the real-time web browsing session, the alternative transaction is associated with at least a threshold probability of having the alternative transaction to be completed by the user, wherein the machine learning model accounts for time when the user previously completed transactions that are fixed at a checkout page based on the frequency with which certain data were included in the prior data transmission transactions, and time when the prior data transmission transactions were previously received;
- identify, in real time during the real-time web browsing session, a second transaction account that is associated with the alternative transaction identified by the machine learning model, the alternative transaction associated with a second data transmission rail, wherein the second transaction account is selected to replace the first transaction account and is associated with a second institution, the second transaction account providing a reward to the user and the reward is increased compared to using the first transaction account to complete the transaction, and wherein identifying the second transaction account comprises:
- receiving information describing the data that are fixed at the checkout page;
- receiving reward program information of a plurality of candidate transaction accounts;
- and
- using the information describing the data that are fixed at the checkout page and the reward program information to select one of the candidate transaction accounts;
- present, by the online system and in real time during the real-time web browsing session, the second transaction account associated with the second data transmission rail to the user;
- receive, by the online system in real time during the real-time web browsing session, a selection of the second transaction account from the user;
- store, at the online system, an association between an identifier of the user, the second transaction account, and the alternative transaction;
- transmit, in real time during the real-time web browsing session, information describing the second transaction account from the online system to the second institution via the second data transmission rail that replaces the first data transmission rail.