Employee Retention Predictor

This is a Bayesian model that predicts employee retention and analyzes potential reasons for turnover based on various factors including workload, salary, and tenure.

Features

• Predicts expected length of employment
• Calculates probability of employee leaving
• Analyzes most likely reason for departure (if applicable)
• Provides confidence intervals for predictions
• Uses Bayesian inference for robust uncertainty quantification

Prerequisites

• Python 3.8 or higher
• Dependencies listed in requirements.txt

Installation

1. Clone this repository
2. Create a virtual environment (recommended):

python -m venv venv
source venv/bin/activate # On Windows use: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

Usage

The model requires two CSV files:

1. Historical data file for training (see data.example.csv)
2. Current employees file for prediction (see current_employees.example.csv)

Running the Model

python run.py --learning data.csv --input current_employees.csv --output report.csv

Input File Formats

Historical Data (learning) CSV

Must include the following columns:

• employee_id: Unique identifier (email or employee ID)
• months_in_company: Duration of employment
• workload_quota: Percentage of workload achievement (e.g., 0.95 = 95%)
• salary_ratio: Employee's salary as percentage of reference salary (e.g., 1.10 = 110%)
• stayed: Boolean (1 = still employed, 0 = left)
• termination_reason: Code for why employee left (0 = N/A, 1 = underperformance, 2 = better offer)

Current Employees CSV

Must include:

• employee_id: Unique identifier (email or employee ID)
• months_in_company: Current duration of employment
• workload_quota: Current workload achievement
• salary_ratio: Current salary ratio

Output Report

The output CSV will contain all input columns plus:

• employee_id: Employee identifier from input
• predicted_longevity: Expected total months of employment
• longevity_ci_50_low/high: 50% confidence interval
• longevity_ci_75_low/high: 75% confidence interval
• probability_of_leaving: Chance of employee leaving (0-1)
• predicted_reason: Most likely reason if leaving
• prob_stays: Probability of staying
• prob_underperformance: Probability of leaving due to underperformance
• prob_better_offer: Probability of leaving for a better offer

Example Files

The repository includes example files:

• data.example.csv: Example historical data
• current_employees.example.csv: Example current employees data

You can test the model using these files:

python run.py --learning data.example.csv --input current_employees.example.csv --output report.csv

License

This project is licensed under the BSD 2-Clause License - see the LICENSE file for details.