Using Machine Learning to Predict G/G/s Queue Performance

A comprehensive analysis using Neural Networks, Random Forest, and XGBoost on simulation-based datasets for queue prediction and optimization.

About the Project

This research project investigates the application of machine learning techniques to predict performance metrics in G/G/s queue systems. By analyzing simulation-generated datasets, the study compares the effectiveness of Neural Networks, Random Forest, and XGBoost algorithms in forecasting queue behavior and system performance.

The project demonstrates how modern machine learning approaches can provide accurate predictions for complex queue systems, offering insights into system optimization and resource allocation strategies.

3
ML Algorithms
2
Queue Models

Project Details

Neural Networks

Multi-layer perceptron implementation with backpropagation for queue prediction, featuring adaptive learning rates and dropout regularization.

  • Multi-layer architecture
  • Backpropagation training
  • Dropout regularization
  • Mean squared error loss

Random Forest

Ensemble method using multiple decision trees with bootstrap sampling and feature importance ranking for robust predictions.

  • Bootstrap sampling
  • Feature importance analysis
  • Out-of-bag validation
  • Ensemble averaging

XGBoost

Gradient boosting implementation with regularization parameters and early stopping to prevent overfitting.

  • Gradient boosting
  • Regularization (L1/L2)
  • Early stopping
  • Cross-validation

Datasets & Models

Queue Simulation Data

The project analyzes performance metrics from two distinct queue system simulations:

M/M/S Queue

Queue with general arrival and service time distributions

G/G/S Queue

Multi-server queue system with parallel processing capabilities

Model Testing

Input Parameters

Range: 0.5 - 0.99
Range: 1 - 20
Range: 1 - 10
Must be greater than λ/s

Prediction Results (Wq)

Neural Network

-
Confidence: -

Random Forest

-
Confidence: -

XGBoost

-
Confidence: -

Key Features & Results

Algorithm Performance

Comparative analysis of prediction accuracy across Neural Networks, Random Forest, and XGBoost models.

Prediction Accuracy

Queue waiting time (Wq) prediction with error analysis.

Contact & Information

Author

Aashrith Raj Tatipamula

Institution

King's University College

Program

King's Internship Program

Project Type

Machine Learning Research