Predictive Customer Retention: A Telecom Solution

Predictive.Customer.Retention-.A.Telecom.Solution.mp4

Overview

Telecom companies face significant challenges with customer churn, often losing subscribers due to dissatisfaction. This project aims to provide a novel solution for proactively identifying and addressing customer dissatisfaction using data-driven and AI-powered techniques.

---

Features

1. Data Synthesis

• Comprehensive dataset incorporating relevant customer features:
  • Call/Network history
  • Usage patterns
  • Billing details
  • Customer feedback

2. ML Model Training

• The XGBoost model was trained on the synthesized dataset.
• The model identifies patterns of customer dissatisfaction.

3. Feature Interpretability (SHAP)

• SHAP (SHapley Additive exPlanations) was used to:
  • Explain the predictions made by the ML model.
  • Identify the most influential features contributing to dissatisfaction for each customer.

4. Strategy Generation

• A fine-tuned Language Model (LLM) generates personalized retention strategies based on the influential feature.
• LLM used: llama3-8b-8192 accessed through groqcloud

5. Automated Actions

• AI-generated strategies are automatically integrated into the customer’s user interface.
• The interface provides personalized recommendations and tailored solutions.

---

Benefits

Personalized Strategies

• Provides targeted offers, discounts, or service upgrades based on dissatisfaction features.

Proactive Communication

• Enables customer service representatives to address specific concerns proactively.

Service Improvement

• Insights generated inform service improvements and policy changes to address underlying dissatisfaction issues.

---

Implementation Details

Schedule Periodic Batch Jobs / Trigger Events

• The system operates during periodic intervals or upon trigger events such as:
  • Customer complaints
  • Feedback forms

Future Directions: Enhancing the strategies using reinforcement learning

Feedback Loop

• Customer engagement feedback is sent back to the AI agent. Whether the strategy worked or not.

Refinement of Strategies

• The feedback is fed back to the LLM (AI system) and it can continuously learn from feedback and data to refine existing strategies.

---

Getting Started

Prerequisites

• Python 3.8 or later
• Required libraries:
  • XGBoost
  • SHAP
  • Transformers

Installation

1. Clone the repository.

   git clone https://github.com/shruthimohan03/Predictive-Customer-Retention-A-Telecom-Solution.git

2. Run the application.

   cd final-UI
   python app.py

---

Contributors

• Shruthi Mohan (GitHub)
• Bagiya Lakshmi (GitHub)

---