An end-to-end Machine Learning pipeline and interactive dashboard designed to identify high-potential talent, predict employee burnout, and optimize workforce performance using Explainable AI (XAI).
Modern HR departments rely on reactive data (yearly reviews, exit interviews) to manage talent. TalentPredict AI transforms this process into a proactive, predictive model. By ingesting employee KPIs (projects completed, training scores, overtime), this system utilizes a Random Forest Classifier to predict performance outcomes and generate actionable, prescriptive business insights.
Fig 1: Main Dashboard featuring KPI tracking, Financial Risk Alerts, and Departmental Filtering.
- 🧠 Predictive Analytics: Uses historical data to classify employees as "High Performers" or "At-Risk," allowing for early intervention.
- 💡 Prescriptive Insights Engine: Goes beyond descriptive charts by auto-generating textual recommendations (e.g., "Schedule upskilling workshops," "Review workload distribution").
- 💸 Financial Risk Modeling: Calculates the estimated dollar-value replacement cost of employees flagged for severe burnout risk.
- ⚖️ Explainable AI (XAI): Features a "Drivers of Performance" module that extracts Gini importance, ensuring algorithmic transparency for HR leadership.
- 🧪 Scenario Simulator: Allows stakeholders to test "What-If" scenarios (e.g., boosting company-wide training scores by 15%) and immediately see the simulated impact on high-performer ratios.
- 📋 Local Explainability: Includes an Employee Deep-Dive tool to compare individual metrics against the company average via interactive Radar Charts.
Fig 2: Explainable AI Feature Importance and Dynamic Workload Distribution Analysis.
Fig 3: Conditional gradient employee roster and Scenario Simulation engine.
The project follows industry-standard modular architecture to separate the ingestion, logic, and UI layers.
Employee-Performance-Predictor/
│
├── src/
│ ├── ingestion.py # Data validation and health scoring
│ ├── insights.py # Logic for prescriptive text generation
│ └── model_engine.py # OOP-based Scikit-Learn training and XAI wrapper
│
├── outputs/ # Directory for screenshots and generated reports
├── app.py # Streamlit UI & Application Core
├── generate_sample_data.py # Script to generate synthetic HR datasets
├── requirements.txt # Dependency management
└── README.md # Project documentation
---
## 🧭 How to Use the Dashboard
Once the Streamlit app is running in your browser, follow these steps to explore the AI engine:
1. **Upload Data:** Look at the left sidebar under "Control Center." Click **Browse files** and upload the `hr_sample_data.csv` file generated in the previous step.
2. **Set the AI Threshold:** Use the slider in the sidebar to adjust the "High Performer Probability Threshold" (e.g., set it to 0.65). Watch how the KPIs dynamically update based on the AI's confidence levels.
3. **Filter by Department:** Use the dropdown in the sidebar to isolate specific teams (like "Sales" or "IT") to see department-specific burnout risks.
4. **Explore Prescriptive Analytics:** Navigate to the **"💡 AI Recommendations"** tab to read the automated executive briefing.
5. **Run a Simulation:** In the Recommendations tab, check the **Scenario Simulator** box to see how a hypothetical 15% increase in training scores would impact the overall high-performer ratio.
6. **Conduct an Employee Deep Dive:** Go to the **"📋 Employee Roster & Deep Dive"** tab. Scroll to the bottom, select a specific `Employee_ID`, and analyze their individual radar chart against the company average.
7. **Export Results:** Click the **"📥 Export Filtered Analytics Report"** button to download the AI-scored roster for downstream HR systems.