Customer Segmentation based Customized Product Recommendation using Unsupervised Learning and LLMs

Customer Segmentation based Customized Product Recommendation using Unsupervised Learning and LLMs

The Challenge

A financial institution aimed to group customers based on diverse characteristics such as demographics, behaviors, hobbies etc. The goal was to identify various customer segments to understand their unique traits. To address this challenge, Tatras was tasked with creating a solution using GenAI and Machine Learning. The system segments customers into different sub-groups based on the selected group and provides detailed descriptions of each sub-group’s characteristics using GenAI.

Hypothesis

  • GenAI can perform feature selection for specific groups eg., selecting features important for demographic groups.
  • ML clustering algorithms can segment subgroups within selected group. Subgroups would be subgroup of selected group.
  • GenAI can generate brief summaries for each subgroup using descriptive statistics.

Execution

  • A GenAI module selects important features for the specified demographic group.
  • A clustering pipeline then used the extracted features to create subgroups within the selected group.
  • A GenAI-based descriptor was implemented to describe each subgroup/ cluster.
  • Libraries used: Scikit-learn, Langchain, Transformers.

Outcomes

  • Obtained the subgroups of a selected group along with their descriptions.
  • This enables the financial institution to tailor its strategies for improved customer engagement.
  • Customized product recommendations that meet the specific needs and preferences of each subgroup.
  • Automated customer segmentation enables the tailoring of strategies for enhanced customer engagement with greater precision in near real-time.

Project Highlights

  • Real-time customer segmentation for dynamic marketing deployed.