🧠 Employee Attrition Analysis & Prediction
📌 Project Objective This project aims to analyze employee attrition behavior and predict employees who are at risk of leaving the company.
Goals:
Identify key factors influencing employee attrition
Detect high-risk employee groups
Generate data-driven HR (Human Resources) insights
Predict attrition probability using machine learning
📊 Technologies Used
Python
Pandas / NumPy → Data manipulation and analysis
DuckDB → SQL-based data analysis
Matplotlib → Data visualization
Scikit-learn → Machine learning modeling
📂 Dataset
The project uses a dataset containing various employee-related features:
Demographic information (Age, MaritalStatus, etc.)
Job-related data (JobRole, Department, BusinessTravel)
Performance & satisfaction metrics (JobSatisfaction, WorkLifeBalance)
Compensation (MonthlyIncome)
Experience (TotalWorkingYears, YearsWithCurrManager)
Target variable: Attrition (Yes / No)
🔍 1. Exploratory Data Analysis (EDA)
SQL queries (via DuckDB) were used to explore the dataset.
📌 General Analysis
Average employee age
Employee count by department
Age distribution by education field
📌 Attrition Analysis
Attrition rate by business travel frequency
Attrition across age groups
Impact of overtime on attrition
Effect of marital status
📌 Key Insights
Employees working overtime have higher attrition rates
Low work-life balance is strongly associated with attrition
Longer time since last promotion increases attrition risk
Younger employees tend to have higher attrition rates
⚙️ 2. Data Preprocessing
Before modeling, the dataset was prepared as follows:
Converted Attrition to binary format (Yes → 1, No → 0)
Dropped irrelevant columns
Categorical variables:
Transformed using One-Hot Encoding
Data split:
Train/Test split (80% / 20%)
Feature scaling:
Applied StandardScaler
🤖 3. Modeling
Model Used:
Logistic Regression
Model Configuration:
class_weight="balanced" → to handle class imbalance
max_iter=2000
📈 4. Model Performance
Evaluation metrics:
Accuracy
ROC AUC Score
Classification Report:
Precision
Recall
F1-score
Additional Visualizations:
Confusion Matrix
ROC Curve
🔎 5. Feature Importance (Interpretability)
Model coefficients were analyzed to understand feature impact:
🔺 Factors Increasing Attrition Risk:
Overtime
Low job satisfaction
Long time since last promotion
Low income
🔻 Factors Decreasing Attrition Risk:
High job satisfaction
Higher income
Strong/long-term manager relationships
Using the trained model:
Current employees were evaluated
High-risk individuals (attrition probability > 70%) were identified
This enables:
Proactive HR interventions
Reduction of employee turnover
💡 Use Case Scenario
HR teams can leverage this model to:
Answer: “Who is likely to leave?”
Design targeted retention strategies
Optimize salary, promotion, and workload policies