Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ConvLSTMTransNet: A Hybrid Deep Learning Approach for Internet Traffic Telemetry

Sajal Saha, Saikat Das, Glaucio H. S. Carvalho

TL;DR

ConvLSTMTransNet addresses internet traffic forecasting by merging CNN-based spatial feature extraction, LSTM-based temporal modeling, and Transformer-based long-range dependency capture. The model is evaluated on real traffic data from provider-edge router interfaces and benchmarked against RNN, LSTM, and GRU baselines using MAE, RMSE, and WAPE, reporting significant accuracy gains. The main contributions include a novel hybrid architecture, a data preprocessing and time-lag feature pipeline, and empirical evidence of improved prediction accuracy on high-speed traffic telemetry. The findings highlight the practical potential of hybrid spatial-temporal architectures for network resource management and real-time traffic forecasting, with future work on multi-variate and online learning and robustness to adversarial perturbations.

Abstract

In this paper, we present a novel hybrid deep learning model, named ConvLSTMTransNet, designed for time series prediction, with a specific application to internet traffic telemetry. This model integrates the strengths of Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and Transformer encoders to capture complex spatial-temporal relationships inherent in time series data. The ConvLSTMTransNet model was evaluated against three baseline models: RNN, LSTM, and Gated Recurrent Unit (GRU), using real internet traffic data sampled from high-speed ports on a provider edge router. Performance metrics such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Weighted Absolute Percentage Error (WAPE) were used to assess each model's accuracy. Our findings demonstrate that ConvLSTMTransNet significantly outperforms the baseline models by approximately 10% in terms of prediction accuracy. ConvLSTMTransNet surpasses traditional models due to its innovative architectural features, which enhance its ability to capture temporal dependencies and extract spatial features from internet traffic data. Overall, these findings underscore the importance of employing advanced architectures tailored to the complexities of internet traffic data for achieving more precise predictions.

ConvLSTMTransNet: A Hybrid Deep Learning Approach for Internet Traffic Telemetry

TL;DR

ConvLSTMTransNet addresses internet traffic forecasting by merging CNN-based spatial feature extraction, LSTM-based temporal modeling, and Transformer-based long-range dependency capture. The model is evaluated on real traffic data from provider-edge router interfaces and benchmarked against RNN, LSTM, and GRU baselines using MAE, RMSE, and WAPE, reporting significant accuracy gains. The main contributions include a novel hybrid architecture, a data preprocessing and time-lag feature pipeline, and empirical evidence of improved prediction accuracy on high-speed traffic telemetry. The findings highlight the practical potential of hybrid spatial-temporal architectures for network resource management and real-time traffic forecasting, with future work on multi-variate and online learning and robustness to adversarial perturbations.

Abstract

In this paper, we present a novel hybrid deep learning model, named ConvLSTMTransNet, designed for time series prediction, with a specific application to internet traffic telemetry. This model integrates the strengths of Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and Transformer encoders to capture complex spatial-temporal relationships inherent in time series data. The ConvLSTMTransNet model was evaluated against three baseline models: RNN, LSTM, and Gated Recurrent Unit (GRU), using real internet traffic data sampled from high-speed ports on a provider edge router. Performance metrics such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Weighted Absolute Percentage Error (WAPE) were used to assess each model's accuracy. Our findings demonstrate that ConvLSTMTransNet significantly outperforms the baseline models by approximately 10% in terms of prediction accuracy. ConvLSTMTransNet surpasses traditional models due to its innovative architectural features, which enhance its ability to capture temporal dependencies and extract spatial features from internet traffic data. Overall, these findings underscore the importance of employing advanced architectures tailored to the complexities of internet traffic data for achieving more precise predictions.
Paper Structure (19 sections, 6 equations, 2 figures, 1 table)

This paper contains 19 sections, 6 equations, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Proposed Methodology
  4. Data Handling and Normalization
  5. Time-Lagged Feature Extraction in Time Series Forecasting
  6. Baseline Models
  7. Recurrent Neural Networks (RNN)
  8. Long Short-Term Memory (LSTM)
  9. Gated Recurrent Units (GRU)
  10. Proposed ConvLSTMTransNet Model Architecture
  11. Input and Convolutional Layer
  12. LSTM Layer
  13. Transformer Encoder Layer
  14. GlobalAveragePooling1D and Output Layer
  15. Experiment and Result Analysis
  16. ...and 4 more sections

Figures (2)

  • Figure 1: Proposed ConvLSTMTransNet Model Architecture Illustrating its Layers and Connections
  • Figure 2: Comparative Visualization of Actual vs. Predicted Traffic Using the ConvLSTMTransNet Model