Network Traffic Prediction using Machine Learning

🚀 IMPORTANT

This repository supports two different workflows:

Method 1 — Wireshark Traffic Prediction

Use this if you have:

• .pcapng files
• Wireshark packet captures
• Wireshark text exports

Method 2 — Kaggle Dataset Training

Use this if you downloaded datasets such as:

• CIC-IDS2017
• UNSW-NB15
• KDD Cup 1999
• Network Traffic Flows

➡️ If you are using a Kaggle dataset, skip directly to Method 2.

---

🤗 Hosted Model

The trained model from this project is published on Hugging Face — you can browse or download it directly instead of retraining from scratch.

---

Features

Wireshark Workflow

• Analyze real network captures
• Parse .pcapng and Wireshark exports
• Generate traffic statistics
• Predict future traffic metrics
• Generate reports and visualizations

Kaggle Workflow

• Train ML models from network datasets
• Automatic dataset column mapping
• Forecast future traffic
• Save reusable models
• Generate feature importance charts

---

Repository Structure

Network-Traffic-Prediction-using-Machine-Learning/
│
├── run.py
├── traffic_predictor.py
├── visualize.py
├── generate_sample_data.py
│
├── kaggle_train.py
├── kaggle_predict.py
│
├── requirements.txt
└── README.md

---

METHOD 1 — Wireshark Traffic Prediction

Use this workflow if you have real packet capture files.

Supported Formats

Format	Description
.pcapng	Wireshark capture
.txt	Wireshark plain text export

---

Installation

pip install -r requirements.txt

---

Run Using PCAPNG

python run.py capture.pcapng

---

Run Using Wireshark Text Export

python run.py capture.txt

---

Demo Mode

python run.py

---

Available Options

--window N

Aggregation window in seconds.

--model-dir

Directory for trained models.

--output-dir

Directory for reports and charts.

--no-plots

Disable graph generation.

---

Output Files

output/
├── prediction_output.txt
├── prediction_report.json
└── plots/
    ├── dashboard.png
    ├── protocol_distribution.png
    └── traffic_timeseries.png

---

Wireshark Export Instructions

1. Open capture in Wireshark
2. File → Export Packet Dissections
3. Select "As Plain Text"
4. Enable "Packet Bytes"
5. Save as .txt

---

Wireshark Prediction Outputs

Throughput / Bandwidth

Example:

At 3:00 PM tomorrow,
the incoming traffic will be 1.2 Gbps.

Packet Count

Example:

We expect 15,000 packets per second
during the next prediction window.

Link Utilization

Example:

The backbone link will be at 78%
capacity between T1 and T2.

---

METHOD 2 — Kaggle Dataset Training

⚠️ If you downloaded a dataset from Kaggle, start here.

---

Recommended Datasets

CIC-IDS2017 (Recommended)

https://www.kaggle.com/datasets/cicdataset/cicids2017

UNSW-NB15

https://www.kaggle.com/datasets/mrwellsdavid/unsw-nb15

Network Traffic Flows

https://www.kaggle.com/datasets/jsrojas/ip-network-traffic-flows-labeled-with-87-apps

KDD Cup 1999

https://www.kaggle.com/datasets/galaxyh/kdd-cup-1999-data

---

Download Dataset

Option A — Manual Download

1. Open dataset page
2. Click Download
3. Extract ZIP
4. Place CSV file beside kaggle_train.py

---

Option B — Kaggle API

Install Kaggle:

pip install kaggle

Create API token:

https://www.kaggle.com/settings

Download dataset:

kaggle datasets download -d cicdataset/cicids2017 --unzip

---

Train Models

Basic training:

python kaggle_train.py --file dataset.csv

Example:

python kaggle_train.py --file Friday-WorkingHours.csv

Custom model directory:

python kaggle_train.py --file dataset.csv --model-dir kaggle_models

Limit rows:

python kaggle_train.py --file dataset.csv --rows 500000

---

Models Trained

Three independent Gradient Boosting models are created:

Model	Prediction
Model 1	Throughput/Bandwidth (Mbps)
Model 2	Packet Count (pkt/s)
Model 3	Link Utilization (%)

---

Training Output

kaggle_models/
├── kaggle_throughput_mbps_model.pkl
├── kaggle_throughput_mbps_scaler.pkl
├── kaggle_packet_count_model.pkl
├── kaggle_packet_count_scaler.pkl
├── kaggle_link_util_pct_model.pkl
├── kaggle_link_util_pct_scaler.pkl
├── kaggle_training_metrics.json
└── feature_importance_plots

---

Predict Future Traffic

Interactive mode:

python kaggle_predict.py --model-dir kaggle_models

Predict using new CSV:

python kaggle_predict.py \
    --model-dir kaggle_models \
    --file new_data.csv

Custom forecast:

python kaggle_predict.py \
    --model-dir kaggle_models \
    --steps 20 \
    --window 60

Disable graph:

python kaggle_predict.py \
    --model-dir kaggle_models \
    --no-plot

---

Kaggle Prediction Outputs

Throughput / Bandwidth

At 3:00 PM tomorrow,
the incoming traffic will be X Mbps/Gbps.

Packet Count

We expect X,XXX packets per second
during the next prediction window.

Link Utilization

The backbone link will operate at X% capacity.

---

Machine Learning Pipeline

Raw Network Data
       │
       ▼
Feature Engineering
       │
       ▼
Traffic Metrics
       │
       ▼
Gradient Boosting Regressor
       │
       ├── Throughput Model
       ├── Packet Count Model
       └── Link Utilization Model
       │
       ▼
Future Traffic Forecast
       │
       ▼
Reports + Visualizations

---

Technologies Used

• Python
• Pandas
• NumPy
• Scikit-Learn
• Matplotlib
• Joblib

---

Quick Start

For Wireshark Users

pip install -r requirements.txt

python run.py capture.pcapng

---

For Kaggle Users

pip install -r requirements.txt

python kaggle_train.py --file dataset.csv

python kaggle_predict.py --model-dir kaggle_models

---