Systems and methods for assessing trustworthiness of customer accounts in omnichannel retail transactions using machine learning

A computer-implemented method, comprising:

• receiving, from a database, original features comprising transactions labelled with a trust level;
• inputting the original features to an encoder of at least one machine learning model to generate encoded features at a bottleneck layer of the at least one machine learning model, wherein the encoder compresses the original features from a first dimensionality to at least a second dimensionality to generate the encoded features, and wherein the second dimensionality is less than the first dimensionality;
• inputting the encoded features to a decoder of the machine learning model to generate reconstructed features, wherein parameters of the machine learning model are adjusted based on a loss between the reconstructed features and the original features;
• receiving, from a computing device over a network, a trust assessment request regarding a transaction for a customer;
• in response to the trust assessment request, receiving, from the database, historical transaction data and relevant account information of the customer;
• generating at least one trust-related feature associated with the customer based on the historical transaction data and the relevant account information of the customer, wherein the at least one trust-related feature comprises a first feature labelled as trusted and a second feature labelled as non-trusted;
• execute the at least one machine learning model and generate trust score data of the customer based on the at least one trust-related feature, wherein the trust score data characterizes a probability that the customer is trustworthy;
• and
• transmitting, in response to the trust assessment request and over the network, the trust score data of the customer to the computing device to process the transaction in accordance with the trust score data.

A non-transitory computer readable medium having instructions stored thereon, wherein the instructions, when executed by at least one processor, cause at least one device to perform operations comprising:

• receiving, from a database, original features comprising transactions labelled with a trust level;
• inputting the original features to an encoder of at least one machine learning model to generate encoded features at a bottleneck layer of the at least one machine learning model, wherein the encoder compresses the original features from a first dimensionality to at least a second dimensionality to generate the encoded features, and wherein the second dimensionality is less than the first dimensionality;
• inputting the encoded features to a decoder of the machine learning model to generate reconstructed features, wherein parameters of the machine learning model are adjusted based on a loss between the reconstructed features and the original features;
• receiving, from a computing device over a network, a trust assessment request regarding a transaction for a customer;
• in response to the trust assessment request, receiving, from the database, historical transaction data and relevant account information of the customer;
• generating at least one trust-related feature associated with the customer based on the historical transaction data and the relevant account information of the customer, wherein the at least one trust-related feature comprises a first feature labelled as trusted and a second feature labelled as non-trusted;
• execute the at least one machine learning model and generate trust score data of the customer based on the at least one trust-related feature, wherein the trust score data characterizes a probability that the customer is trustworthy;
• and
• transmitting, in response to the trust assessment request and over the network, the trust score data of the customer to the computing device to process the transaction in accordance with the trust score data.