Layoff Risk Predictor

A Machine Learning project that predicts layoff percentages of companies using data from Layoffs.fyi. This project compares multiple regression models — from simple linear regression to advanced ensemble techniques — to identify key economic and organizational factors influencing workforce downsizing.

---

Repository

🔗 GitHub: https://github.com/praventhegenius/Layoff-Risk-Predictor.git

---

Overview

Objective

To predict company layoff percentages based on multiple parameters such as:

• Company funding stage
• Industry sector
• Geographical location
• Funds raised
• Year and month of operation

Motivation

Layoffs have surged globally, creating economic instability and uncertainty. This project aims to assist organizations, researchers, and policymakers in proactively identifying at-risk companies or sectors through data-driven modeling.

---

Methodology

1. Data Collection

• Source: Layoffs.fyi Dataset (2020–2024)
• Format: CSV
• Size: ~3,000 records post-cleaning

2. Preprocessing

• Removed duplicates and handled missing values
• Encoded categorical features (industry, stage, country)
• Scaled numeric attributes (funds_raised)
• Chronological data split: Training (≤2023), Testing (>2023)

3. Modeling

Implemented and compared:

• Linear Regression
• Ridge Regression
• Lasso Regression
• Elastic Net Regression
• KNN Regressor
• Decision Tree Regressor
• Random Forest Regressor
• CatBoost Regressor
• XGBoost Regressor

Each model was tuned using cross-validation for hyperparameter optimization.

---

Visualizations

The repository includes plots for:

• Actual vs Predicted Layoff %
• Residual and Q–Q Plots
• Feature Importance (Tree-Based Models)
• MAE vs Hyperparameter Curves
• Decision Tree Visualization

Example:

Actual vs Predicted Layoff % (CatBoost)
Feature Importance (Stage, Industry, Funds Raised)
Residuals vs Predicted (Decision Tree)
MAE vs Alpha (Ridge/Lasso)

---

Model Comparison

Model	Category	MAE ↓	R² ↑	Key Insight
Linear Regression	Linear	0.058	0.41	Baseline model
Ridge	Linear (L2)	0.051	0.47	Stable coefficients
Lasso	Linear (L1)	0.049	0.50	Sparse feature selection
Elastic Net	Hybrid	0.047	0.52	Balanced regularization
KNN	Non-Parametric	0.045	0.56	Captures local trends
Decision Tree	Tree	0.042	0.60	Interpretable splits
Random Forest	Ensemble	0.037	0.68	Strong generalization
CatBoost	Boosted Ensemble	0.034	0.72	Best overall performer

---

Key Takeaways

• Stage, industry, and funds raised are the most predictive factors for layoffs.
• CatBoost Regressor offered the best combination of accuracy and interpretability.
• Ensemble models outperform single regressors due to non-linear layoff behavior.

---

Future Work

• Integrate macroeconomic indicators (inflation, GDP, unemployment)
• Deploy interactive dashboards for real-time layoff risk prediction
• Explore neural architectures (RNN, Transformer) for temporal layoff forecasting

---

Tech Stack

• Language: Python
• Libraries: pandas, numpy, scikit-learn, matplotlib, seaborn, catboost
• Tools: Jupyter Notebook, GitHub, Visual Studio Code

---

How to Run

1. Clone this repository

   git clone https://github.com/praventhegenius/Layoff-Risk-Predictor.git
   cd Layoff-Risk-Predictor

2. Install dependencies

   pip install -r requirements.txt

3. Run the notebook

   jupyter notebook Layoff_Risk_Predictor.ipynb

4. View results

   • Visualizations under /visuals
   • Model comparison outputs in terminal or notebook cells

---

Author

Pravenraam Shankar : GitHub Profile – praventhegenius

Tanmay Sapra : GitHub Profile – tanmaysapra

---

🏁 License

This project is open-sourced under the MIT License — feel free to use, modify, and cite with attribution.

---