HR Attrition
Predictor

This study uses machine learning to predict and understand employee attrition, aiming to provide actionable insights for proactive human resource strategies based on diverse employee attributes.

This project focuses on addressing the challenge of employee attrition within organizations through a comprehensive analysis using machine learning techniques. The study employs Logistic Regression, Random Forest, Support Vector Machines (SVM), and Decision Trees to forecast and understand instances of employee turnover. Leveraging historical data encompassing a wide range of employee attributes, from demographics to nuanced factors like job role satisfaction, the project aims to construct accurate predictive models while unraveling the complex interplay of influential factors contributing to workforce turnover.

Key Components:

Exploratory Data Analysis (EDA): In-depth analysis of the IBM HR Analytics Employee Attrition & Performance dataset using Python, aiming to uncover insights into the factors influencing employee attrition within the organization.

Machine Learning Models: Implementation and evaluation of Logistic Regression, Random Forest, SVM, and Decision Trees to predict employee attrition. Each model's performance is assessed using metrics such as accuracy, precision, recall, and F1-score.

Dashboard: Visualization of the analysis results through Power BI, providing interactive reports and insights to facilitate decision-making for HR professionals.

Findings and Recomendations

• Identification of top factors contributing to attrition, including Monthly Income, Age, and Over Time.
• Comparison of machine learning model performances, with insights into strengths and limitations of each approach.
• Recommendations for proactive human resource management strategies based on the analysis, such as age-specific approaches, compensation reviews, and addressing overtime concerns.

Model Comparison

• Logistic Regression: Achieved an accuracy of 86.73%, exhibiting moderate precision and recall. Struggled to predict positive instances accurately.
• Random Forest: Outperformed with an accuracy of 87.75%, displaying higher precision and recall. Identified most instances correctly but misclassified some positive cases.
• Support Vector Machine (SVM): Mirrored Logistic Regression's metrics, achieving 86.73% accuracy. Similar challenges in predicting positive instances.
• Decision Tree: Accuracy of 86.05%, with a moderate balance between precision and recall. Effective overall correctness but room for improvement in predicting positive cases.

Conclussion

The project concludes with actionable insights and recommendations aimed at helping organizations mitigate employee attrition and foster a stable, engaged, and retention-focused work environment.