NanoMQTTMessageTransformer

A collection of MQTT message transformation, monitoring, and analytics examples built with .NET.

This repository explores techniques for:

• MQTT message ingestion and republishing
• Message transformation pipelines
• Dynamic C# script-based transformations
• MQTT loopback testing
• Device liveness monitoring
• Spike detection
• Change-point detection
• ML.NET-based analytics for MQTT telemetry

Solution Structure

The repository contains several sample projects focused on different MQTT processing scenarios. [github.com]

src/
├── HiveMQClient/
├── IMessageTransformer/
├── LightGBMBinaryClassificationTrainer/
├── LightGBMDataGeneratorEvT/
├── LightGBMMultiClassClassificationTrainer/
├── LightGBMRegressionTrainer/
├── MQTTChangePointDetection/
├── MQTTCLientCodeLoopback/
├── MQTTChangePointDetection/
├── MQTTCLientCodeLoopback/
├── MQTTClientCSScriptDynamicLoadingLoopback/
├── MQTTClientCSScriptLoopback/
├── MQTTDeviceSimulator/
├── MQTTLightGBM/
├── MQTTLivenessMonitor/
├── MQTTSpikeDetection/
├── MQTTSRCNNSpikeDetection/
├── NFMQTTUltrasonicRanger/
└── MQTTMachineLearningWithMLNET.slnx

Key Features

Dynamic Message Transformation

Use C# scripts to transform MQTT payloads without rebuilding applications.

Anomaly Detection

Detect unusual telemetry patterns including:

• Sudden spikes
• Sensor drift
• Change points in streaming data

Device Monitoring

Monitor MQTT-connected devices and identify offline or unresponsive nodes.

Machine Learning Integration

Experiment with ML.NET models for real-time telemetry analysis and event detection.

Example Use Cases

• IoT sensor monitoring
• Edge-to-cloud telemetry processing
• MQTT topic routing
• Sensor anomaly detection
• Industrial monitoring
• Rapid prototyping of MQTT transformations

Architecture

MQTT Publisher
       │
       ▼
  MQTT Broker
       │
       ▼
 Message Consumer
       │
       ▼
 Transformation Engine
       │
       ├── Rule-Based Processing
       ├── C# Script Processing
       ├── Spike Detection
       ├── Change Detection
       └── ML.NET Analysis
       │
       ▼
 MQTT Publisher / Alerts / Storage

Warning

• These projects are intended as experiments and learning exercises.
• Error handling, security, and production readiness may vary between PoCs.

Background

• Message Transformation in code with HiveMQ Client

• Message Transformation with cached transform binaries

• Real-time IoT Spike and Change Point Detection with ML.NET

• Real-time IoT Change Point Detection with ML.NET

• Real-time Spike Detection with SR-CNN

• nanoFramework RS485 Light Intensity Sensor

• nanoFramework RS485 Temperature, Humidity & Dewpoint Sensor

• nanoFramework RS485 Temperature, Humidity & CO2 Sensor

• nanoFramework RS485 500cm Ultrasonic Level Sensor

Blog posts will be published once I build a couple more nanoFramework apps to collect some training data *

• Real-time ML.Net Regression with LightGBM

• Real-time ML.Net Binary classification with LightGBM

• Real-time ML.Met Multi class classification with LightGBM

The MQTT examples should be considered as "snappy" rather than "real-time". For embedded/edge compute "real-time" is different often with specified values for up-time, duration, latency and jitter.

License

MIT License

---