Utilizing a machine learning model to determine attribution for communication channels

A device, comprising:

• one or more memories;
• and
• one or more processors, communicatively coupled to the one or more memories, configured to:
• receive customer data identifying touchpoints of customers with an entity,
• wherein the customer data is data associated with one or more marketing channels or one or more marketing campaigns;
• identify unique communication channels associated with the touchpoints identified in the customer data;
• determine, based on the customer data, an optimal order for a Markov chain model to be utilized to determine attribution weights for the unique communication channels;
• determine a model accuracy of the Markov chain model based on the optimal order;
• transform transitions in the Markov chain model, based on the optimal order and the customer data, to generate transformed transitions;
• train a multi-level indexing model to establish relationships between the customer data based on the unique communication channels and the transformed transitions;
• utilize the multi-level indexing model to establish relationships between the customer data based on the unique communication channels and the transformed transitions;
• generate sparse matrices based on the relationships between the customer data,
• wherein the sparse matrices include at least a portion of the customer data;
• determine removal effects and steady state values for the sparse matrices;
• determine attribution weights for the unique communication channels based on the Markov chain model with the optimal order, the removal effects, and the steady state values;
• and
• perform one or more actions associated with customer sales or customer conversion based on the attribution weights,
• wherein the one or more processors, when performing the one or more actions, are to:
• cause marketing plans to be implemented for the customers based on the attribution weights;
• and
• retrain the multi-level indexing model based on the attribution weights.

A non-transitory computer-readable medium storing instructions, the instructions comprising:

• one or more instructions that, when executed by one or more processors, cause the one or more processors to:
• receive customer data identifying touchpoints of customers with an entity,
• wherein the customer data is data associated with one or more marketing channels or one or more marketing campaigns;
• identify unique communication channels associated with the touchpoints identified in the customer data;
• determine, based on the customer data, an optimal order for a Markov chain model to be utilized to determine attribution weights for the unique communication channels;
• determine a model accuracy of the Markov chain model based on the optimal order;
• transform transitions in the Markov chain model, based on the optimal order and the customer data, to generate transformed transitions;
• train a multi-level indexing model to establish relationships between the customer data based on the unique communication channels and the transformed transitions;
• process the customer data, with the multi-level indexing model and based on the unique communication channels and the transformed transitions, to generate sparse matrices,
• wherein the sparse matrices include at least a portion of the customer data;
• utilize a linear system of equations to calculate removal effects and steady state values for the sparse matrices;
• determine attribution weights for the unique communication channels based on the Markov chain model with the optimal order, the removal effects, and the steady state values;
• and
• perform one or more actions associated with customer sales or customer conversion based on the attribution weights,
• wherein the one or more instructions, that cause the one or more processors to perform the one or more actions, cause the one or more processors to at least:
• determine marketing plans for the customers based on the attribution weights, and
• retrain the multi-level indexing model based on the attribution weights.