Proactive Measures of the Organization Regarding Employee Attrition Using Deep Learning
Author: SINGAMPALLI RAMASRI1 (MCA student), M. BALA NAGA BHUSHANAMU2 (Asst. Prof) Department of Information Technology & Computer application,
Andhra University College of Engineering, Visakhapatnam, AP.
Corresponding Author: SINGAMPALLI RAMASRI
(email-id: singampalliramasri@gmail.com)
ABSTRACT: Employee attrition remains a persistent concern for organizations, often leading to increased costs and disruptions in workforce planning. To address this issue, we present a predictive system that leverages deep learning to assess the likelihood of employee resignation. The proposed solution combines a PyTorch-based Multi-Layer Perceptron (MLP) with a Flask-powered web interface, enabling real-time, user-friendly attrition prediction. The model is trained on structured HR data that includes both numerical and categorical attributes related to employee demographics, job roles, and workplace behaviour. Preprocessing steps involve one-hot encoding for categorical variables and normalization for numerical inputs, ensuring compatibility with the neural network. The application calculates a probability score using sigmoid activation applied post-inference and classifies the outcome based on a configurable threshold. The web interface collects user inputs and delivers immediate feedback on attrition risk, making the system accessible to HR personnel without technical expertise. This approach demonstrates the feasibility of integrating machine learning models into operational HR workflows and highlights the role of deep learning in improving employee retention strategies. The system is modular, scalable, and suitable for extension to include additional factors such as employee feedback or performance metrics.
Keywords: Keywords Employee Attrition, Deep Learning, Multi-Layer Perceptron, Predictive Analytics, Human Resource Management, PyTorch, Flask, Feature Engineering, Classification, Workforce Analytics
