Personalized next-best action recommendation with multi-party interaction learning for automated decision-making

Area of Science:

• Decision Science
• Artificial Intelligence
• Machine Learning

Background:

• Personalized next-best action recommendation is essential in dynamic, interactive contexts for business and social decision-making.
• Existing models struggle to quantify complex, multi-sequence interactions involving customer states, behaviors, and reactions.

Purpose of the Study:

• To develop a data-driven approach for personalized next-best action recommendation.
• To address limitations of current modeling theories in capturing intricate decision-making processes.

Main Methods:

• Utilized a reinforced coupled recurrent neural network (CRN) for personalized decision-making.
• CRN models multiple coupled dynamic sequences of customer states, responses, and rewards.
• Learned long-term, multi-sequence interactions between customers and decision-makers.

Main Results:

• CRN effectively quantifies complex, personalized decision-making dynamics.
• Demonstrated the capability to recommend next-best actions to optimize customer states.
• Showcased automated dynamic intervention for improved decision outcomes.

Conclusions:

• Personalized deep learning of multi-sequence interactions offers a powerful approach for automated decision-making.
• The CRN model provides a novel solution for complex, interactive decision systems.
• This data-driven method enhances personalized interventions in dynamic environments.