International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 04 | Apr 2025
p-ISSN: 2395-0072
www.irjet.net
Optimized Nurse Scheduling and Patient Admission Prediction Using Machine Learning Pranav Shukla1, Atharva Jadhav2, Sakshi Mahure3, Disha Shete 4, Prof. Pramila M. Chawan 5 1B.Tech Student, Dept of Computer Engineering and IT, VJTI College, Mumbai, Maharashtra, India 2B.Tech Student, Dept of Computer Engineering and IT, VJTI College, Mumbai, Maharashtra, India 3B.Tech Student, Dept of Computer Engineering and IT, VJTI College, Mumbai, Maharashtra, India
4B.Tech Student, Dept of Computer Engineering and IT, VJTI College, Mumbai, Maharashtra, India 5Associate Professor, Dept of Computer Engineering, VJTI College, Mumbai, Maharashtra, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Efficient nurse scheduling is vital for hospital
management to optimize workforce utilization and maintain fairness. This paper presents a Hybrid Ensemble Model, combining Genetic Algorithm (GA), Ant Colony Optimization (ACO), Particle Swarm Optimization (PSO), Simulated Annealing (SA), Reinforcement Learning (RL), and forecasting models (XGBoost/LSTM) to generate adaptive schedules. We also introduce a Modified PDE-ODE model for patient demand forecasting using neural networks with time series features. The combined system enhances scheduling efficiency, balances workloads, and improves responsiveness to dynamic patient needs.
●
In addition to scheduling optimization, this study incorporates a Predictive Demand Forecasting Model powered by Partial Differential Equations (PDE) and Ordinary Differential Equations (ODE). This model uses PyTorch and deep learning techniques to predict future hospital admissions, enabling more accurate staffing decisions. The PDE-ODE framework enhances the scheduling system by providing reliable patient demand predictions, which improves resource allocation and reduces staff overburdening.
Key Words: Nurse Scheduling Problem, Patient Demand Forecasting, Ensemble Method, Reinforcement Learning, Workforce Optimization, Hospital Management, Machine Learning
1.INTRODUCTION
The PDE-ODE model processes historical hospital admissions data with lag features, moving averages, and temporal indicators (e.g., day of the year, week of the year). It employs a neural network architecture with batch normalization, dropout layers, and early stopping to optimize accuracy and prevent overfitting. The model achieves reliable R² and RMSE scores, demonstrating its effectiveness in capturing complex temporal patterns. By integrating this forecasting model into the scheduling system, the proposed solution dynamically adjusts nurse allocations based on predicted patient influx, leading to better resource management and improved patient care.
Efficient nurse scheduling is a critical aspect of hospital management, aiming to balance workforce constraints, patient care demands, and employee satisfaction. The Nurse Scheduling Problem (NSP) is a combinatorial optimization challenge that requires assigning nurses to shifts while adhering to complex hard and soft constraints. Traditional scheduling methods, such as rule-based algorithms and standalone metaheuristics like Genetic Algorithm (GA) or Reinforcement Learning (RL), often struggle with real-time adaptability, fairness, and workload balancing. These limitations can lead to inefficient scheduling, overburdened staff, and compromised patient care.
This integrated framework enhances the fairness, adaptability, and efficiency of nurse scheduling systems in dynamic hospital environments, offering a comprehensive solution for workforce optimization and demand forecasting.
To overcome these challenges, we introduce a Hybrid Ensemble Model for Nurse Scheduling that combines metaheuristic optimization with reinforcement learning and forecasting techniques. Our model employs a multiphase pipeline: ● Global Search Phase: Utilizes GA and Ant Colony Optimization (ACO) to generate diverse initial schedules and reinforce promising solutions. ● Local Refinement Phase: Leverages Particle Swarm Optimization (PSO) and Simulated
© 2025, IRJET
|
Impact Factor value: 8.315
Annealing (SA) to fine-tune the schedule, minimizing constraint violations and balancing workloads. Dynamic Adjustment Phase: Uses reinforcement learning (RL) and a forecasting model (XGBoost/LSTM) to adapt schedules based on real-time changes, such as nurse absences or fluctuating patient demand.
2. LITERATURE SURVEY Efficient nurse scheduling is vital for hospital management to balance staffing constraints and ensure quality patient care. Traditional scheduling methods often fail to handle
|
ISO 9001:2008 Certified Journal
|
Page 129