LearnDash AI

An AI-Powered Learner Retention Dashboard

Live App [Public link]: https://huggingface.co/spaces/vshnnuu/LearnDash-AI

LearnDash is an AI-powered Dashboard that uses an end-to-end machine learning system for predicting learner churn/unsubscription risk for an online education platform and recommends retention strategies.
The project combines a predictive model, a decision agent, and a simulated CRM workflow inside an interactive dashboard.

---

Overview

Online learning platforms often lose learners due to inactivity, low engagement, billing issues, or unresolved support problems.

LearnDash's system predicts churn probability and generates targeted interventions such as:

• retention discount offers
• personalized course recommendations
• proactive support outreach
• re-engagement email campaigns

The goal is to demonstrate how ML predictions can integrate with operational decision workflows.

---

System Architecture

User Input (Dashboard)
        ↓
Churn Prediction Model
        ↓
Risk Classification
        ↓
Retention Decision Agent
        ↓
Intervention Strategy
        ↓
CRM Workflow Simulation

---

Major pipeline features

Preprocessing

• Missing value imputation
• Feature scaling for numeric features
• One-hot encoding for categorical features

Implemented using scikit-learn pipelines and ColumnTransformer.

Retention Decision Agent

A rule-based agent analyzes learner signals and model output to determine:

Churn Drivers

• long inactivity
• low learning activity
• payment failures
• unresolved support issues
• low engagement

Retention Strategies

• discount offers
• reactivation campaigns
• course recommendations
• proactive support outreach

The agent prioritizes actions based on signal severity.

CRM Simulation Layer

To mimic real operational workflows, the system simulates CRM actions such as:

[COMPLETED] Discount offer created — OFF-1042
[QUEUED] Email campaign queued — EML-2208
[OPEN] Support follow-up task created — SUP-3314
[QUEUED] Learning recommendation workflow queued — LRN-5586

This demonstrates how predictive systems connect with marketing and customer success tools in production environments.

Interactive Dashboard

The system includes a Gradio dashboard where users can:

• input learner attributes
• simulate engagement behavior
• predict churn probability
• view recommended retention actions
• see simulated CRM task execution

Inputs are grouped into tabs such as:

• Learner Profile
• Subscription Details
• Learning Activity
• Engagement Signals
• Billing & Marketing

Running the Project

1. Create environment (for setting up locally)

python -m venv .venv
source .venv/bin/activate

2. Install dependencies

pip install -r requirements.txt

3. Train the model

python -m src.train

4. Launch the dashboard

python app.py

The interface will start at:

http://127.0.0.1:7860

Tech Stack

• Python
• scikit-learn
• XGBoost
• Pandas
• Gradio
• Joblib