Systems and methods for determining user personas for a user

A computer-implemented method comprising:

• receiving, by a processor, a persona request identifying a first user and a first platform;
• receiving, by the processor, user session data representative of interactions between the first user and the first platform occurring during a concurrent session;
• obtaining, by the processor, historical user data associated with the user from a database, the user data including user session data identifying and characterizing interactions between the user and the platform and user transaction data identifying and characterizing one or more transactions of the user;
• obtaining, by the processor, a plurality of potential user personas from the database, each of the plurality of potential user personas including data characterizing a particular product category of a plurality of product categories;
• generating one or more embeddings based on the user session data, the historical user data, and the plurality of potential user personas;
• training a first model to generate a platform-specific sub-score, wherein the first model is trained using at least the one or more embeddings generated based on the historical user data;
• training a second model to generate a user-specific sub-score, wherein the second model is trained using at least the one or more embeddings generated based on the historical user data and the plurality of potential user personas;
• for each of the plurality of potential user personas, determining, by the processor, a combination score for the user based on at least the user session data and each of the plurality of potential user personas, the combination score indicates affinity to a corresponding potential user persona for the first platform, wherein the combination score is generated by aggregating outputs of the first and second models;
• selecting, by the processor, at least one potential user persona of the plurality of potential user personas as a final user persona for the user and the platform based on the a corresponding combination score;
• and
• transmitting, by the processor and over one or more networks, the final user persona to a computing system configured to (i) generate item recommendation data identifying and characterizing one or more item recommendations based on the final user persona and (ii) transmit, over the one or more networks and to a second computing device, instructions configured to cause the second computing device to generate an interface including the one or more item recommendations;
• receiving additional session data representative of one or more additional interactions between the first user and the first platform in response to generation of the interface;
• generating one or more updated embeddings representative of the additional session data;
• and
• retraining at least one of the first model or the second model based on the one or more embeddings and the one or more updated embeddings.

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:

• receiving a persona request identifying a first user and a first platform;
• receiving user session data representative of interactions between the first user and the first platform occurring during a concurrent session;
• obtaining historical user data associated with the user from a database, the user data including user session data identifying and characterizing interactions between the user and the platform and user transaction data identifying and characterizing one or more transactions of the user;
• obtaining a plurality of potential user personas from the database, each of the plurality of potential user personas including data characterizing a particular product category of a plurality of product categories;
• generating one or more embeddings based on the user session data, the historical user data, and the plurality of potential user personas;
• training a first model to generate a platform-specific sub-score, wherein the first model is trained using at least the one or more embeddings generated based on the historical user data;
• training a second model to generate a user-specific sub-score, wherein the second model is trained using at least the one or more embeddings generated based on the historical user data and the plurality of potential user personas;
• for each of the plurality of potential user personas, determining a combination score for the user based on at least the user session data and each of the plurality of potential user personas, the combination score indicates affinity to a corresponding potential user persona for the first platform, wherein the combination score is generated by aggregating outputs of the first and second models;
• selecting at least one potential user persona of the plurality of potential user personas as a final user persona for the user and the platform based on a corresponding combination score;
• and
• transmitting, over one or more networks, the final user persona to a computing system configured to (i) generate item recommendation data identifying and characterizing one or more item recommendations based on the final user persona and (ii) transmit, over the one or more networks and to a second computing device, instructions configured to cause the second computing device to generate an interface including the one or more item recommendations;
• receiving additional session data representative of one or more additional interactions between the first user and the first platform in response to generation of the interface;
• generating one or more updated embeddings representative of the additional session data;
• and
• retraining at least one of the first model or the second model based on the one or more embeddings and the one or more updated embeddings.